|Home | About | Journals | Submit | Contact Us | Français|
To investigate whether variation in the distribution of the risk allele frequency of the Y402H single-nucleotide polymorphism (SNP) across various ethnicities and geographic regions reflects differences in the prevalence of late age-related macular degeneration (AMD) in those ethnicities.
Published data were obtained via a systematic search. Study samples were grouped into clusters by ethnicity and geographic location and the Spearman correlation coefficient of the prevalence of late AMD and risk allele frequencies was calculated across clusters.
Across all ethnicities, AMD prevalence was seen to increase with age. Populations of European descent had both higher risk allele frequencies and prevalence of late AMD than did Japanese, Chinese, and Hispanic descendants. Results for African descendants were anomalous: although allele frequency was similar to that in European populations, the age-specific prevalence of late AMD was considerably lower. The correlation coefficient for the association between allele frequency and AMD prevalence was 0.40 (95% confidence interval [CI] = −0.36 to 0.84, P = 0.28) in all populations combined and 0.71 (95% CI = 0.02–0.94, P = 0.04) when people of African descent were excluded.
Evidence was found at the population level to support a positive association between the Y204H risk allele and the prevalence of AMD after exclusion of studies undertaken on persons of African ancestry. Data in African, Middle Eastern, and South American populations are needed to provide a better understanding of the association of late AMD genetic risk across ethnicities.
Age-related macular degeneration (AMD) is the leading cause of blindness in older people in Western populations. Several lifestyle and environmental risk factors have been identified for AMD,1 and more recently, genetic variants with strong effects have also been identified.2 The genetic variants have mainly been studied in people of European origin. In this study, we focused on the single-nucleotide polymorphism (SNP) Y402H in the complement factor H (CFH) gene, which is characterized by a substitution of histidine for tyrosine at codon 402 on the long arm of chromosome 1, region 31 (rs1061170). This SNP is particularly striking because of the strength of its association with late AMD. Odds ratios (ORs) >5 were found for those homozygous for the Y402H risk allele, making this genetic association one of the strongest for a complex disorder yet to be reported.3
Although most of the population-based studies reporting the prevalence of late AMD have been undertaken in people of European origin in Western settings, studies are now emerging from other geographic areas. These, along with studies in different ethnic subgroups within Western populations (principally the United States) suggest that there is considerable variation in the distribution of risk alleles of the Y402H SNP4–8 and that these variations may in part explain the differences in the prevalence of late AMD in the respective population groups. To our knowledge, no studies have been conducted to examine whether differences in the frequency of this variant between populations or ethnic groups can explain differences in the prevalence of late AMD. We therefore conducted an ecological correlation of the prevalence of late AMD with the frequencies of the Y402H risk allele across ethnic groups and geographic regions. We considered only late AMD because of the differences in classification and grading of early AMD across studies.
We systematically searched the literature for studies of the prevalence of late AMD and studies that included representative data on the frequency of Y402H genotypes or alleles in the general population.
We searched PubMed for relevant articles using the search terms “prevalence” in combination with “age-related macular degeneration” or “AMD,” limiting our search to English language papers. The articles that we deemed eligible were those that clearly confirmed the use of robust systems of fundus grading to determine the diagnosis of AMD9,10 and reported prevalence of late AMD (either neovascular, geographic atrophy, or a combination of both), by age groups and ethnicity of the sample. The search identified 22 eligible articles, some of which reported prevalence in samples from two or more ethnic groups, resulting in a total of 29 unique samples. The studies reported AMD prevalence using different age ranges starting from as low as 35 to above 90.
We searched PubMed for relevant articles by using the search terms “age-related macular degeneration” and “gene” or “complement factor H” as well as “complement factor H” on its own. We searched for articles on association studies of AMD as well as non-AMD diseases with the Y402H polymorphism. Among these, we included those that reported SNP allele or genotype frequencies of non-AMD cases. Studies were included irrespective of whether Hardy-Weinberg equilibrium (HWE) was observed for the Y402H genotypes in non-AMD cases, since this is not necessarily reflective of poor study design, but may instead be reflective of copy number variation within the complement factor genetic region.11 Furthermore, in studies in which HWE was observed, the association between the Y402H risk allele and AMD has been shown to be the same as in studies in which HWE was not observed (see, for example, Thakkinstian et al.12). Our search resulted in 50 eligible articles, some of which included two or more samples or samples from more than one ethnic group, hence 67 unique samples. Tables 1 and and22 give general characteristics of the prevalence and Y402H studies that were selected in our search.
Our hypothesis was that the observed differences in the Y402H risk allele frequencies between ethnicities reflects various prevalence rates of late AMD in these ethnicities. We assumed that the estimates given by the samples in the studies are measures of true allele frequencies and prevalence of late AMD in their respective populations and geographic regions. We thus defined clusters of samples within studies based on both ethnicity and geographic region. For example, people of non-Hispanic European origin living in the United States were categorized in the European-American cluster. For each cluster, mean log-transformed prevalence and mean risk allele frequencies were calculated across all samples, weighted by the sample sizes. Spearman's rank correlation coefficient was calculated.
The prevalence of late AMD and allele frequency data in our analysis came from independent samples within ethnicities and regions, except for four studies that sampled from the same populations: MVIP (Melbourne Visual Impairment Project), LALES (Los Angeles Latino Eye Study), ProyectoVER (Vision and Eye Research Project), and the Rotterdam Eye Study (Tables 1 and and2).2). Our selected publications produced 11 clusters on prevalence of AMD and 12 on Y402H studies. Of these, nine clusters had data on both the prevalence and Y402H frequencies. The age groups used in the prevalence studies varied slightly from study to study, and we therefore merged some of these to create three common groups: below 60, between 60 and 69, and above 70 years.
In the studies on prevalence of late AMD, a clear increase was observed with advancing age but there was a marked variation between studies in the prevalence rates. Most notably, lower rates were observed in Chinese, Japanese, and African Americans. At the other extreme were very high prevalence rates in the Icelandic population, whereas the Greenland Inuit had an overall prevalence that was even 2.5 times higher than that of the Icelandics (Table 3, Fig. 1).
In studies on CFH Y402H allele distribution, low-risk allele frequencies (<1%) were observed in the Japanese, Chinese, and Koreans. In most other populations studied, the allele frequencies were between 36% and 42% (Table 4, Figs. 2, ,3).3). Figure 4 shows a plot of the log transformed, weighted AMD prevalence by age group against the weighted risk allele frequencies. This relationship, across all age groups, is plotted in Figure 5. Only nine clusters are depicted on the graphs because other clusters (Inuit Greenland, South African, Malay, Korean, Israeli, and Japanese-Brazilians) either had prevalence or allele frequency data, but not both, as shown in Tables 3 and and44.
Spearman's rank correlation between weighted AMD prevalence and CFH risk allele frequency was 0.40 (P = 0.28) and after the studies undertaken on persons of African ancestry were excluded, it was 0.71 (P = 0.04).
Overall, our results suggest evidence of a positive association between the prevalence of late AMD and Y402H risk-allele frequency across ethnicities, except in those of African descent. We observed marked differences in both AMD prevalence and allele frequency in different ethnicities and geographic regions. We therefore used an ecological study design to combine evidence of gene–disease association based on published data. To our knowledge, until now, this approach has not been applied to the study of AMD. It provides a useful way of comparing gene–disease associations between ethnicities and geographic regions when large interpopulation-based studies do not exist. The criteria we used to select studies ensured that we included prevalence estimates that were measured using standard diagnostic techniques in different ethnicities. We did not analyze the results by type of late AMD, because not all studies reported prevalence in that way, and therefore we used estimates of any late AMD.
There are several limitations in our analysis. Only four of the prevalence studies listed in Table 1 had the risk allele frequency derived from the same populations. Most of the studies on the Y402H variant were designed as case–control studies and therefore AMD prevalence was not established within them. Risk allele frequencies were strongly homogeneous within each of the clusters, and therefore it was sensible to cluster study samples in that way. This homogeneity within clusters suggests that, at least for the studies in our analysis, within a given homogenous residential population, a representative sample of genetic information is meaningful for others in the same population. With increased ethnic mixing, populations may be more genetically heterogeneous, and self-reported ethnicity may not reflect this. Ethnic mixing and migration may apply to populations of African origin, but the results of the three African American studies and one South African study show similar allele frequencies.
It is interesting that the prevalence of late AMD among African descendants does not match their allele frequencies. This mismatch is unlikely to be due to bias in the corresponding prevalence studies, as they have been carefully designed and rigorously implemented. It may be that an association of the Y402H risk allele with late AMD does not exist or is less pronounced in these populations. At present, there are not sufficient data from association studies on those of African ancestry. There is only one study in Africa in which the association between Y204H and AMD (early form) was investigated.13 The investigators found a nonsignificant odds ratio of 1.56 (95% confidence interval [CI] = 0.75–3.33) for the risk allele, but it was based on a very small sample size. A neighboring deletion allele delCFHR1 confers a protective effect in Europeans.14 Based on high frequencies of Y402H and delCHFR1 alleles in African Americans, the hypothesis in one study15 of the complement factor H and related genes was that the effect of these explains the low prevalence of late AMD in African Americans. This notion is in line with what our ecological analysis showed. Further studies are needed in African populations to substantiate this conclusion. It is also plausible that differences between Africans and Europeans in risk factors such as smoking and sunlight may partly explain the lower prevalence of AMD in people of African descent, but at present, there is insufficient evidence to address this point.
There is a notably high prevalence of late AMD among the Icelandics and the Greenland Inuit. Only one prevalence study was available from each of these two fairly environmentally similar but racially different regions, with sample sizes of 922 and 660 for Iceland and Greenland, respectively. The prevalence estimates in these studies may be imprecise due to small sample sizes, but the response rates for both studies were high (76% and 75%), and so the estimates are unlikely to be biased. The allele frequency for the Icelandic study was 0.38, which is comparable to those of European populations. There may be environmental and lifestyle factors that contribute to such a high prevalence of late AMD in these two populations.
Differences in observed prevalences within a given cluster may reflect the paucity of people in the very elderly age group in some studies, perhaps due to response bias. (Older people may be less likely to participate for reasons associated with AMD.) Differences between clusters may also have arisen because, in some ethnicities, the proportion of the older age group is lower in the population. A low number of people in the oldest age groups, where AMD rates are the highest, can lead to imprecision in the estimation of the prevalence (i.e., sampling error). More detailed data, such as the mean ages within the age categories, would have enabled us to investigate potential age bias between ethnicities.
We could not include environmental risk factors that are known to be associated with AMD, such as tobacco smoking,16,17 because most papers reporting the association of these factors with AMD do not mention the prevalence of the risk factors for the population controls. It is unlikely, however, that even a strong risk factor such as smoking can explain much of the variation in prevalence of AMD. For example, most studies document a lower prevalence of smoking in women, but there is little, if any, variation in prevalence between men and women. In addition, there is no reason to suspect that allele frequency would be related to such environmental factors. We note that the Y402H SNP has been found to be in linkage disequilibrium with other loci on the CFH gene,18,19 suggesting that there is a haplotype effect. We limited our study to this particular SNP because it is the most widely reported genetic variant in relation to AMD and appears to have very strong effects. We note that there are other genes that increase the risk of development of AMD and that such genes may differ across ethnic groups.
In conclusion, we found evidence at the population level to support a positive association with the Y204H risk allele and the prevalence of AMD. We highlight the anomalous result of the high-risk allele frequency and the low prevalence of late AMD in African Americans. There are none or very few articles on populations from the African continent or indeed from other regions, such as the Middle East, parts of Asia, and South America. In some of these areas, only the proportion of blindness due to AMD in a population has been reported, rather than prevalence of late AMD among the elderly. This deficiency highlights the need for better reporting of findings from these populations obtained with standard diagnostic techniques to quantify the effects of the Y402H SNP (and other risk genetic variants) on AMD prevalence across ethnicities and regions.
Supported by British Heart Foundation (Schillingford) Clinical Training Fellowship FS/07/011 (RS) and by a Wellcome Trust Senior Clinical Fellowship (LS).
Disclosure: B.A.S. Nonyane, None; D. Nitsch, None; J.C. Whittaker, None; R. Sofat, None; L. Smeeth, None; U. Chakravarthy, None; A.E. Fletcher, None