A possible association between apolipoprotein E polymorphisms and age-related macular degeneration has been investigated numerous times, with conflicting results. A previous analysis pooling results from four studies (Schmidt et al., Ophthalmic Genet 2002;23:209–23) suggested an association, but those investigators did not document allele frequencies, the magnitude of the association, or the possible genetic mode of action. Thus, the authors searched MEDLINE from 1966 to December 2005 for any English-language studies reporting genetic associations. Data and study quality were assessed in duplicate. Pooling was performed while checking for heterogeneity and publication bias. Frequencies of the E₂ and E₄ alleles in Caucasians were approximately 8% and 15%, respectively. Allele- and genotype-based tests of association indicated a risk effect of up to 20% for E₂ and a protective effect of up to 40% for E₄. E₂ appeared to act in a recessive mode and E₄ in a dominant mode. There appears to be a differential effect of the E₂ and E₄ alleles on the risk of age-related macular degeneration, although the possibility of survivor bias needs to be ruled out more definitively.

Keywords: ApoE; apolipoproteins E; epidemiology; genetics; macular degeneration; meta-analysis; polymorphism, genetic
Abbreviations: AMD, age-related macular degeneration; ApoE, apolipoprotein E; CI, confidence interval; OR, odds ratio

Age-related macular degeneration (AMD) is the leading cause of blindness in the developed world (1–4), accounting for half of all new cases of registered blindness (5). With an aging population, the burden of AMD is set to grow, with almost 30 percent of persons aged 75 years or older showing early signs of disease (6–8). The pathologic hallmark of the disease is drusen, deposits of protein and lipid, in the retinal pigment epithelium or Bruch's membrane. This maculopathy progresses to degeneration in two forms: 1) geographic atrophy, in which there is loss of retinal pigment epithelium and photoreceptors, and 2) neovascular AMD, in which there is choroidal neovascularization and hemorrhages.

Little is known about the pathogenesis of AMD. Smoking is the only established risk factor, although other cardiovascular disease risk factors (e.g., high cholesterol, hypertension) may also play a role (9). There also appears to be a genetic component, as supported by a number of lines of evidence: familial aggregation (10–13), segregation analysis (14, 15), twin studies (10, 16, 17), and several linkage studies (18–24) culminating in a meta-analysis (25). Although several monogenic forms of macular dystrophy have been described and their genes identified (for reviews, see Yates et al. (15) and Gorin et al. (26)), these have not shed light on sporadic AMD.

Apolipoprotein E (ApoE) is a lipid transport protein that acts as a ligand for the low density lipoprotein receptor, but it is also involved in the repair and maintenance of neuronal cell membranes in the central and peripheral nervous system. The ApoE gene is located at chromosome 19q13.2, and three major forms, originally identified by isoelectric focusing, have been described. These isoforms are defined by amino acid changes at positions 112 and 158: Alleles E₂, E₃, and E₄ are defined respectively by cysteine/cysteine, cysteine/arginine, and arginine/arginine at these two sites (27, 28).

ApoE is involved in clearance of chylomicrons and very low density lipoproteins from the circulation via specific receptors on liver and peripheral cells. The E₂ form of ApoE has decreased affinity for the receptor, whereas the E₃ and E₄ forms have higher affinity. Mutations in ApoE lead to type III hyperlipoproteinemia, in which there is an increase in triglycerides and cholesterol and premature cardiovascular disease (29). ApoE has also been linked to Alzheimer's disease. Since 1993, when Saunders et al. (30) reported an increased frequency of the E₄ allele in a small prospective series of Alzheimer's disease patients visiting a memory disorders clinic, multiple studies have confirmed an increased risk of Alzheimer's disease among persons with the E₄ allele.

The genetic association between ApoE and AMD has been investigated multiple times, although the results have been inconsistent (4, 8, 31–36). A previous study summarizing results from four groups (8) suggested a protective effect of the E₄ allele and a risk effect of the E₂ allele in comparison with the most common allele, E₃. We pooled the results of all available population-based studies of the association between ApoE and AMD to ascertain whether there is a genetic effect on AMD susceptibility and, if so, to estimate the magnitude of that effect and the possible genetic mode of action (37, 38).

Search strategy
We searched MEDLINE (US National Library of Medicine) for all relevant articles published from January 1966 through November 23, 2005, using the PubMed search engine. The search strategy was "macular degeneration" and "apolipoprotein E*" or "apoE" or "APOE" or "Apo E." Results were limited to English-language papers.

Inclusion criteria
Any human population-based association study, regardless of sample size, was included if it met the following criteria (we use the term "population-based" to refer to individual sporadic cases rather than familial cases or family-based study designs (e.g., sibling pairs)):

• The investigators determined the association between the ApoE polymorphism and AMD. The alleles and genotypes for this polymorphism were, respectively: E₂, E₃, and E₄; and E₂E₂, E₂E₃, E₂E₄, E₃E₃, E₃E₄, and E₄E₄.
  
• The outcome was AMD and there were at least two comparison groups (e.g., AMD vs. control (non-AMD) groups). For those studies in which AMD was graded (i.e., drusen, pigment abnormalities in retinal pigment epithelium, geographic atrophy, and choroidal neovascularization), these gradings were collapsed into only one AMD group.
  
• There were sufficient results for extraction of data (i.e., the number of subjects with each genotype in the AMD and control groups).
  

Data extraction
Data were extracted independently and in duplicate by two reviewers (T. A. and B. S.) using a standardized data extraction form. Data on covariables such as mean age, gender, and ethnicity were also extracted for each study. Any disagreement was adjudicated by a third author (A. J.).

Quality score assessment
The quality of the studies was independently assessed by two reviewers (B. S. and M. M.) using a quality assessment score developed for genetic association studies (39). This score was based on both traditional epidemiologic considerations and genetic issues (40). Total scores ranged from 0 (worst) to 12 (best). Any disagreement was adjudicated by a third author (T. A.).

Statistical analysis
Hardy-Weinberg equilibrium was assessed for each study using the chi-squared test (41–43). The summary prevalence of all alleles was estimated and characterized using only the data on controls (39). Both per-allele analysis and per-genotype analysis were performed.

Per-allele analysis.
The association between ApoE polymorphisms and AMD was first determined using the per-allele approach. Allele frequencies were calculated for studies reporting only genotype data.

The Q test for heterogeneity was performed separately for two odds ratios (ORs), that is, E₂ versus E₃ (OR₁) and E₄ versus E₃ (OR₂). Logistic regression analysis was used to determine the overall gene effect. Bivariate meta-analysis with the maximum likelihood method was used to estimate the summary odds ratios (44–46). OR₁ and OR₂ and their 95 percent confidence intervals are reported.

Per-genotype analysis.
Once a gene effect was confirmed, per-genotype analysis was used to ascertain the genetic model. Two separate per-genotype analyses—E₂E₂, E₂E₃, and E₃E₃ and E₄E₄, E₃E₄, and E₃E₃—were separately explored by assigning E₃E₃ as the reference group. The genotype effects were estimated using the model-free approach (47), in which no assumptions about genetic models are required. OR₃ (E₂E₂ vs. E₃E₃), OR₄ (E₂E₃ vs. E₃E₃), OR₅ (E₄E₄ vs. E₃E₃), and OR₆ (E₃E₄ vs. E₃E₃) were estimated using multivariate meta-analysis with Bayesian methods. The log odds ratios were modeled accounting for both between- and within-study variation. Two separate lambda values (reflecting the genetic model), which were the ratios of log OR₄ to log OR₃ for ₁ and log OR₆ to log OR₅ for ₂, were estimated. These parameters capture information about the genetic mode of action, as follows: The model is a recessive model if = 0, a dominant model if = 1, a codominant model if = 0.5, and a homozygous or heterosis model if > 1 or < 0.

For per-allele and per-genotype analyses, we took two approaches for handling Hardy-Weinberg disequilibrium. Firstly, we performed sensitivity analyses by including and excluding studies not in Hardy-Weinberg equilibrium. Secondly, we included all studies regardless of Hardy-Weinberg equilibrium and instead adjusted for the degree of disequilibrium using the inbreeding coefficient (F) as described by Trikalinos et al. (48). Briefly, for per-allele analyses, the variance of the odds ratio was adjusted by 1 + F (49). Case and control groups were combined for estimation of F using the method described by Ayres and Balding (50). For genotype analysis, predicted genotype frequencies were estimated in control groups (50), and we used the predicted frequencies instead of the observed frequencies in the summary analysis.

We also performed subgroup analysis in Caucasians. Publication bias was assessed using Egger's test (51, 52). In addition, we conducted a cumulative meta-analysis to assess whether the gene effect changed over time (52). All analyses were performed using Stata, version 9.0 (53), except for the per-genotype analysis, which was performed using WinBUGS 1.4.1 (54). For Bayesian modeling, a vague prior distribution, representing the lack of prior information about parameter values (i.e., log odds ratios and ), was specified using normal-distribution priors for both log odds ratios and . A "burn-in" of 10,000 iterations was carried out for the models, followed by 50,000 iterations for parameter estimates. A p value less than 0.05 was considered statistically significant, except for tests of heterogeneity, where a level of 0.10 was used.

Studies identified
Twenty-eight studies were identified by our search strategies. Eighteen of these studies were not eligible (five were not association studies (55–59), five were reviews (3, 8, 15, 26, 60), three were family-based studies (20, 61, 62), two were animal studies (63, 64), one reported only methods (65), one enrolled diabetic subjects with AMD (66), and one was a duplicate (67)), leaving 10 studies (4, 8, 31–36, 68, 69) for inclusion in this analysis. The 10 studies are described in Table 1. Among them, eight studies were carried out in Caucasians and two in Asians. The mean age ranged from 70.9 years to 81.0 years for cases and from 37.0 years to 76.6 years for controls. The percentage of males ranged from 31.6 percent to 63.3 percent.

click to view

Table 1: General characteristics of studies included in a meta-analysis of apolipoprotein E polymorphisms and age-related macular degeneration

Summary prevalences of the E₂ allele were similar for Caucasians and Asians (8.2 percent (95 percent confidence interval (CI): 7.3, 9.0) vs. 9.1 percent (95 percent CI: 6.3, 11.7)) but were more divergent for E₄ (14.9 percent (95 percent CI: 13.8, 16.0) vs. 8.1 percent (95 percent CI: 5.5, 10.6)).

ApoE and AMD
Allele-based methods.
Among the 10 studies included, one did not observe Hardy-Weinberg equilibrium (33) (see Table 2), leaving nine studies (seven of Caucasians and two of Asians) for assessing the association between the ApoE gene and AMD. OR₁ (E₂ vs. E₃) and OR₂ (E₄ vs. E₃) were estimated for each study (Table 2). Neither OR₁ (E₂ vs. E₃) nor OR₂ (E₄ vs. E₃) showed any evidence of heterogeneity (OR₁: ² = 4.50, df = 8, p = 0.81; OR₂: ² = 9.16, df = 8, p = 0.33). Logistic regression indicated that the overall gene effect was significant (likelihood ratio test: 26.39, df = 2, p < 0.01). The summary OR₁ and OR₂, obtained using bivariate meta-analysis, were 1.17 (95 percent CI: 1.01, 1.35) and 0.67 (95 percent CI: 0.57, 0.78), respectively. This means that patients who had an E₂ allele were approximately 17 percent more likely to have AMD than patients with the E₃ allele. Conversely, persons with an E₄ allele were approximately 33 percent less likely to have AMD than persons with allele E₃.

click to view

Table 2: Allele frequencies in studies of apolipoprotein E polymorphisms among patients with age-related macular degeneration and controls

Genotype-based methods.
Table 3 shows the frequencies of the ApoE genotype in case and control groups. We estimated the genotype effects for E₂E₂ and E₂E₃ in each study by assigning the E₃E₃ genotype as the reference group (Table 4). Cells with a zero count had 0.5 added. Two Asian studies (34, 36) did not have E₂E₂ genotypes in either case groups or control groups, and thus results for these studies could not be summarized. The summary odds ratios for the E₂E₂ and E₂E₃ genotypes were 1.05 (95 percent CI: 0.52, 2.20) and 1.22 (95 percent CI: 0.96, 1.61), respectively. These point estimates can be interpreted as meaning that persons with the E₂E₂ and E₂E₃ genotypes had 5 percent and 22 percent higher risks of developing AMD than persons with the E₃E₃ genotype, although these effects did not reach statistical significance. The estimated was 0.27 (95 percent CI: –3.98, 4.93), which suggests a largely recessive mode of action, although the confidence interval was wide.

click to view

Table 3: Genotype frequencies in studies of apolipoprotein E polymorphisms among patients with age-related macular degeneration and controls

click to view

Table 4: Estimation of the summary odds ratios (ORs) OR3 (E2E2 vs. E3E3) and OR4 (E2E3 vs. E3E3) in an analysis of apolipoprotein E polymorphisms among patients with age-related macular degeneration and controls

click to view

Figure 1: Odds ratios (ORs) OR3 (E2E2 vs. E3E3) and OR4 (E2E3 vs. E3E3) in a cumulative meta-analysis of apolipoprotein E polymorphisms among patients with age-related macular degeneration and controls. Horizontal bars, 95% confidence interval.

We also estimated the genotype effects for E₄E₄ and E₃E₄ as compared with E₃E₃ (Table 5). Again, we could not summarize data for the two Asian studies (34, 36), since there was no one with the E₄E₄ genotype in either case groups or control groups in those studies. The pooled OR₅ (E₄E₄ vs. E₃E₃) and OR₆ (E₃E₄ vs. E₃E₃) were 0.85 (95 percent CI: 0.44, 1.75) and 0.62 (95 percent CI: 0.46, 0.90), respectively; that is, persons with the E₄E₄ and E₃E₄ genotypes were approximately 15 percent and 38 percent less likely to have AMD than persons with the E₃E₃ genotype. The estimated was 1.17 (95 percent CI: –4.51, 5.71), which suggests a dominant mode of action.

click to view

Table 5: Estimation of the summary odds ratios (ORs) OR5 (E4E4 vs. E3E3) and OR6 (E3E4 vs. E3E3) in an analysis of apolipoprotein E polymorphisms among patients with age-related macular degeneration and controls

click to view

Figure 2: Odds ratios (ORs) OR5 (E4E4 vs. E3E3) and OR6 (E3E4 vs. E3E3) in a cumulative meta-analysis of apolipoprotein E polymorphisms among patients with age-related macular degeneration and controls. Horizontal bars, 95% confidence interval.

Our study found that the prevalences of ApoE alleles E₂ and E₄ were largely similar between Caucasians and Asians. Assuming a per-allele model, which allowed us to conduct one overall test of association and avoid multiple comparisons, we found that ApoE gene polymorphisms were indeed associated with AMD, with the E₄ allele appearing to be protective and E₂ appearing to be a risk allele. Over half of the studies included (5/9) were of good quality, with a quality score of 10 or above out of 12. Exploring this association in more detail allowed us to estimate the magnitude of the association and the possible genetic mode of action. Results for the E₂ allele did not reach statistical significance, but point estimates appeared to indicate up to a 20 percent increase in risk of AMD and suggested a recessive model. Results for the E₄ allele did reach statistical significance and indicated that E₄ might act dominantly, with the presence of at least one E₄ allele providing up to a 38 percent reduction in the risk of AMD.

These results are strengthened by the facts that there was no evidence of heterogeneity and that including the one study not in Hardy-Weinberg equilibrium gave us similar results. Egger's test evaluates whether small studies produce different results than larger studies. If so, publication bias is a possibility. In this meta-analysis, the result of Egger's test was not significant, but with only 10 studies we had limited power to detect such an effect. The indications of a risk effect of E₂ and a protective effect of E₄ are consistent with results from an earlier, smaller pooling study (8). In addition, the two main types of AMD may have different etiologies (9) and hence may have different genetic susceptibilities. However, there was insufficient information provided in the papers to meta-analyze these two types of AMD separately. Combining both as a single AMD outcome would introduce measurement error in the outcome factor and could lead to a bias towards the null. This bias makes our significant results more robust.

Nevertheless, there are two major concerns. Firstly, this E₄ effect is the opposite of that found for cardiovascular disease; that is, E₄ is associated with increased mortality and decreased longevity. Hence, one might expect that any survivor bias would "deplete" the E₄ allele among persons old enough to develop AMD, and this decreased odds ratio might therefore be spurious. This remains a possibility because, although most studies had similar age distributions in cases and controls, cases had older mean ages than controls. In defense of our results is the finding that all studies but one were in Hardy-Weinberg equilibrium, and the one not in Hardy-Weinberg equilibrium was excluded from the summary analysis. However, it is puzzling that the risk allele for cardiovascular disease should have a beneficial effect for AMD if the mechanism is still related to cholesterol metabolism. This could indicate a type I (i.e., false-positive) error, although there are many examples of pleiotropy in biology (i.e., multiple functions for the same protein or gene), and it is difficult to predict the direction of a genetic effect given the biology. Secondly, the studies included in this meta-analysis were all small or medium-sized case-control studies. There is some evidence that smaller studies tend to overstate genetic effects in comparison with larger studies (70).

Despite these potential problems, our results are statistically robust and point to some interesting directions for future research. In particular, this review indicates the need for confirmation of these results in a large-scale, long-term longitudinal study in which survivor bias might be detected.

Conflict of interest: none declared.

Editor's note: This article is also available on the website of the Human Genome Epidemiology Network (http://www.cdc.gov/genomics/hugenet/default.htm).

• Table 1
  
• Table 2
  
• Table 3
  
• Table 4
  
• Table 5
  
• Figure 1
  
• Figure 2

Appendix Tables

• Appendix Table 1
• Appendix Table 2

1. Mitchell P, Smith W, Attebo K, et al. Prevalence of age-related maculopathy in Australia. The Blue Mountains Eye Study. Ophthalmology 1995;102:1450–60.
2. VanNewkirk MR, Nanjan MB, Wang JJ, et al. The prevalence of age-related maculopathy: The Visual Impairment Project. Ophthalmology 2000;107:1593–600.
3. Pang CP, Lam DS. Differential occurrence of mutations causative of eye diseases in the Chinese population. Hum Mutat 2002;19:189–208.
4. Klaver CC, Kliffen M, van Duijn CM, et al. Genetic association of apolipoprotein E with age-related macular degeneration. Am J Hum Genet 1998;63:200–6.
5. Evans J, Wormald R. Is the incidence of registrable age-related macular degeneration increasing? Br J Ophthalmol 1996;80:9–14.
6. Klein R, Klein B, Linton K. Prevalence of age-related maculopathy: The Beaver Dam Eye Study. Ophthalmology 1992;99:933–43.
7. Vingerling JR, Dielemans I, Hofman A, et al. The prevalence of age-related maculopathy in the Rotterdam Study. Ophthalmology 1995;102:205–10.
8. Schmidt S, Klaver C, Saunders A, et al. A pooled case-control study of the apolipoprotein E (APOE) gene in age-related maculopathy. Ophthalmic Genet 2002;23:209–23.
9. Tomany SC, Wang JJ, Van Leeuwen R, et al. Risk factors for incident age-related macular degeneration: pooled findings from 3 continents. Ophthalmology 2004;111:1280–7.
10. Klein ML, Mauldin WM, Stoumbos VD. Heredity and age-related macular degeneration. Observations in monozygotic twins. Arch Ophthalmol 1994;112:932–7.
11. Seddon J, Ajani U, Mitchell B. Familial aggregation of age-related maculopathy. Am J Ophthalmol 1997;123:199–206.
12. Klaver CC, Wolfs RC, Assink JJ, et al. Genetic risk of age-related maculopathy. Population-based familial aggregation study. Arch Ophthalmol 1998;116:1646–51.
13. Silvestri G, Johnston PB, Hughes AE. Is genetic predisposition an important risk factor in age-related macular degeneration? Eye 1994;8:564–8.
14. Heiba IM, Elston RC, Klein BE, et al. Sibling correlations and segregation analysis of age-related maculopathy: The Beaver Dam Eye Study. Genet Epidemiol 1994;11:51–67.
15. Yates JR, Moore AT. Genetic susceptibility to age related macular degeneration. J Med Genet 2000;37:83–7.
16. Meyers SM, Greene T, Gutman FA. A twin study of age-related macular degeneration. Am J Ophthalmol 1995;120:757–66.
17. Hammond CJ, Webster AR, Snieder H, et al. Genetic influence on early age-related maculopathy: a twin study. Ophthalmology 2002;109:730–6.
18. Klein ML, Schultz DW, Edwards A, et al. Age-related macular degeneration. Clinical features in a large family and linkage to chromosome 1q. Arch Ophthalmol 1998;116:1082–8.
19. Weeks DE, Conley YP, Tsai HJ, et al. Age-related maculopathy: an expanded genome-wide scan with evidence of susceptibility loci within the 1q31 and 17q25 regions. Am J Ophthalmol 2001;132:682–92.
20. Weeks DE, Conley YP, Tsai HJ, et al. Age-related maculopathy: a genomewide scan with continued evidence of susceptibility loci within the 1q31, 10q26, and 17q25 regions. Am J Hum Genet 2004;75:174–89.
21. Majewski J, Schultz DW, Weleber RG, et al. Age-related macular degeneration—a genome scan in extended families. Am J Hum Genet 2003;73:540–50.
22. Schick JH, Iyengar SK, Klein BE, et al. A whole-genome screen of a quantitative trait of age-related maculopathy in sibships from the Beaver Dam Eye Study. Am J Hum Genet 2003;72:1412–24.
23. Seddon JM, Santangelo SL, Book K, et al. A genomewide scan for age-related macular degeneration provides evidence for linkage to several chromosomal regions. Am J Hum Genet 2003;73:780–90.
24. Abecasis GR, Yashar BM, Zhao Y, et al. Age-related macular degeneration: a high-resolution genome scan for susceptibility loci in a population enriched for late-stage disease. Am J Hum Genet 2004;74:482–94.
25. Fisher SA, Abecasis GR, Yashar BM, et al. Meta-analysis of genome scans of age-related macular degeneration. Hum Mol Genet 2005;14:2257–64.
26. Gorin MB, Breitner JC, De Jong PT, et al. The genetics of age-related macular degeneration. Mol Vis 1999;5:29.
27. Weisgraber KH, Rall SC Jr, Mahley RW. Human E apoprotein heterogeneity. Cysteine-arginine interchanges in the amino acid sequence of the apo-E isoforms. J Biol Chem 1981;256:9077–83.
28. Rall SC Jr, Weisgraber KH, Innerarity TL, et al. Structural basis for receptor binding heterogeneity of apolipoprotein E from type III hyperlipoproteinemic subjects. Proc Natl Acad Sci U S A 1982;79:4696–700.
29. National Center for Biotechnology Information, National Library of Medicine. +107741. Apolipoprotein E; APOE. In: OMIM: Online Mendelian Inheritance in Man. (Database). Bethesda, MD: National Library of Medicine, 2006.
30. Saunders AM, Strittmatter WJ, Schmechel D, et al. Association of apolipoprotein E allele epsilon 4 with late-onset familial and sporadic Alzheimer's disease. Neurology 1993;43:1467–72.
31. Schultz DW, Klein ML, Humpert A, et al. Lack of an association of apolipoprotein E gene polymorphisms with familial age-related macular degeneration. Arch Ophthalmol 2003;121:679–83.
32. Baird PN, Guida E, Chu DT, et al. The epsilon2 and epsilon4 alleles of the apolipoprotein gene are associated with age-related macular degeneration. Invest Ophthalmol Vis Sci 2004;45:1311–15.
33. Simonelli F, Margaglione M, Testa F, et al. Apolipoprotein E polymorphisms in age-related macular degeneration in an Italian population. Ophthalmic Res 2001;33:325–8.
34. Gotoh N, Kuroiwa S, Kikuchi T, et al. Apolipoprotein E polymorphisms in Japanese patients with polypoidal choroidal vasculopathy and exudative age-related macular degeneration. Am J Ophthalmol 2004;138:567–73.
35. Souied EH, Benlian P, Amouyel P, et al. The epsilon4 allele of the apolipoprotein E gene as a potential protective factor for exudative age-related macular degeneration. Am J Ophthalmol 1998;125:353–9.
36. Pang CP, Baum L, Chan WM, et al. The apolipoprotein E epsilon4 allele is unlikely to be a major risk factor of age-related macular degeneration in Chinese. Ophthalmologica 2000;214:289–91.
37. Thakkinstian A, McElduff P, D'Este C, et al. A method for meta-analysis of molecular association studies. Stat Med 2005;24:1291–306.
38. Thakkinstian A, D'Este C, Eisman J, et al. Meta-analysis of molecular association studies: vitamin D receptor gene polymorphisms and BMD as a case study. J Bone Miner Res 2004;19:419–28.
39. Thakkinstian A, McEvoy M, Minelli C, et al. Systematic review and meta-analysis of the association between ß2-adrenoceptor polymorphisms and asthma: a HuGE review. Am J Epidemiol 2005;162:201–11.
40. Attia J, Thakkinstian A, D'Este C. Systematic review of meta-analyses of molecular association studies: methodologic lesson for genetic epidemiology. J Clin Epidemiol 2003;56:297–303.
41. Nam J. Simple test for the Hardy-Weinberg law for HLA data with no observed double blanks. Biometrics 1995;51:354–7.
42. Hernandez JL, Weir BS. A disequilibrium coefficient approach to Hardy-Weinberg testing. Biometrics 1989;45:53–70.
43. Chen JJ, Thomson G. The variance for the disequilibrium coefficient in the individual Hardy-Weinberg test. Biometrics 1999;55:1269–72.
44. van Houwelingen HC, Arends LR, Stijnen T. Advanced methods in meta-analysis: multivariate approach and meta-regression. Stat Med 2002;21:589–624.
45. Van Houwelingen HC, Zwinderman KH, Stijnen T. A bivariate approach to meta-analysis. Stat Med 1993;12:2273–84.
46. Thompson JR, Minelli C, Abrams KR, et al. Meta-analysis of genetic studies using Mendelian randomization—a multivariate approach. Stat Med 2005;24:2241–54.
47. Minelli C, Thompson JR, Abrams KR, et al. The choice of a genetic model in the meta-analysis of molecular association studies. Int J Epidemiol 2005;34:1319–28.
48. Trikalinos TA, Salanti G, Khoury MJ, et al. Impact of violations and deviations in Hardy-Weinberg equilibrium on postulated gene-disease associations. Am J Epidemiol 2006;163:300–9.
49. Schaid DJ, Jacobsen SJ. Biased tests of association: comparisons of allele frequencies when departing from Hardy-Weinberg proportions. Am J Epidemiol 1999;149:706–11.
50. Ayres KL, Balding DJ. Measuring departures from Hardy-Weinberg: a Markov chain Monte Carlo method for estimating the inbreeding coefficient. Heredity 1998;80:769–77.
51. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629–34.
52. Egger M, Smith GD, Altman DG. Systematic reviews in health care: meta-analysis in context. London, United Kingdom: BMJ Books, 2001.
53. Stata Corporation. Stata statistical software, release 9.0. College Station, TX: Stata Corporation, 2005.
54. Spiegelhalter D, Thomas A, Best N, et al. WinBUGS user manual. Version 1.4, January 2003. Cambridge, United Kingdom: MRC Biostatistics Unit, Institute of Public Health, 2003.
55. Anderson DH, Ozaki S, Nealon M, et al. Local cellular sources of apolipoprotein E in the human retina and retinal pigmented epithelium: implications for the process of drusen formation. Am J Ophthalmol 2001;131:767–81.
56. Anderson DH, Talaga KC, Rivest AJ, et al. Characterization of beta amyloid assemblies in drusen: the deposits associated with aging and age-related macular degeneration. Exp Eye Res 2004;78:243–56.
57. Ishida BY, Bailey KR, Duncan KG, et al. Regulated expression of apolipoprotein E by human retinal pigment epithelial cells. J Lipid Res 2004;45:263–71.
58. Malek G, Li CM, Guidry C, et al. Apolipoprotein B in cholesterol-containing drusen and basal deposits of human eyes with age-related maculopathy. Am J Pathol 2003;162:413–25.
59. Mullins RF, Russell SR, Anderson DH, et al. Drusen associated with aging and age-related macular degeneration contain proteins common to extracellular deposits associated with atherosclerosis, elastosis, amyloidosis, and dense deposit disease. FASEB J 2000;14:835–46.
60. Kuhrt H, Hartig W, Grimm D, et al. Changes in CD44 and ApoE immunoreactivities due to retinal pathology of man and rat. J Hirnforsch 1997;38:223–9.
61. Kenealy SJ, Schmidt S, Agarwal A, et al. Linkage analysis for age-related macular degeneration supports a gene on chromosome 10q26. Mol Vis 2004;10:57–61.
62. Jun G, Klein BE, Klein R, et al. Genome-wide analyses demonstrate novel loci that predispose to drusen formation. Invest Ophthalmol Vis Sci 2005;46:3081–8.
63. Kliffen M, Lutgens E, Daemen MJ, et al. The APO*E3-Leiden mouse as an animal model for basal laminar deposit. Br J Ophthalmol 2000;84:1415–19.
64. Malek G, Johnson LV, Mace BE, et al. Apolipoprotein E allele-dependent pathogenesis: a model for age-related retinal degeneration. Proc Natl Acad Sci U S A 2005;102:11900–5.
65. Robman L, McNeil J, Dimitrov P, et al. Methodology of the Cardiovascular Health and Age-Related Maculopathy (CHARM) Study. Ophthalmic Epidemiol 2004;11:161–79.
66. Santos A, Salguero ML, Gurrola C, et al. The epsilon4 allele of apolipoprotein E gene is a potential risk factor for the severity of macular edema in type 2 diabetic Mexican patients. Ophthalmic Genet 2002;23:13–19.
67. van Leeuwen R, Klaver CC, Vingerling JR, et al. Cholesterol and age-related macular degeneration: is there a link? Am J Ophthalmol 2004;137:750–2.
68. Schmidt S, Saunders AM, De La Paz MA, et al. Association of the apolipoprotein E gene with age-related macular degeneration: possible effect modification by family history, age, and gender. Mol Vis 2000;6:287–93.
69. Zareparsi S, Reddick AC, Branham KE, et al. Association of apolipoprotein E alleles with susceptibility to age-related macular degeneration in a large cohort from a single center. Invest Ophthalmol Vis Sci 2004;45:1306–10.
70. Ioannidis JP, Trikalinos TA, Ntzani EE, et al. Genetic associations in large versus small studies: an empirical assessment. Lancet 2003;361:567–71.