Estrogen Receptor α Gene Polymorphisms Associated with Incident Aging Macula Disorder

Abstract

purpose. It has been suggested that early menopause increases the risk of aging-macula disorder (AMD), the major cause of incurable blindness with a dry and wet late subtype, and that exposure to endogenous or postmenopausal exogenous estrogens reduces this risk. This study was undertaken to investigate whether genetic variations in the estrogen receptor α (ESR1) gene are associated with incident AMD.

methods. In the Rotterdam Study, a prospective population-based cohort study of participants aged 55 years and older, associations between ESR1 PvuII-XbaI haplotypes and incident early or late AMD were studied in 4571 participants after a mean follow-up time of 7.7 years. Cox proportional hazards regression was used to estimate hazard ratios (HRs) and corresponding 95% confidence intervals (CIs), with adjustment for the most common confounders.

results. ESR1 PvuII-XbaI haplotype 1 was a risk factor for late AMD. Persons with two copies of haplotype 1 were at 3.20 (95% CI, 1.47–6.99) times higher risk for late AMD than noncarriers of haplotype 1, after adjustment for age and sex. This increase was more pronounced for wet AMD (hazard ratio [HR] 4.29; 95% CI, 1.47–12.49) after adjustment for age, sex, smoking, and complement factor H genotype. Correction for additional confounders, including age at menopause, use of hormone replacement therapy, blood pressure, and body mass index did not essentially alter the findings.

conclusions. Persons with one or two copies of ESR1 PvuII-XbaI haplotype 1 have an increased risk of late AMD, especially of the wet form.

Aging macula disorder (AMD), as we now prefer to call age-related macular degeneration, ¹ is a multifactorial disease that affects the center of the retina, the macula lutea, resulting in a central scotoma in the visual field. ^{2 3 4} The pathogenesis of AMD is still not clear, and therapeutic strategies are limited. In our rapidly aging population, with a predicted doubling by 2020 of the persons with blindness due to AMD, it is important to identify risk factors for AMD, to develop possible preventive and therapeutic strategies. 

Diminished exposure to endogenous estrogens as in early menopause has been cross-sectionally associated with AMD, and postmenopausal hormone replacement therapy seems to protect against the disease. ^{5 6 7 8} Although some studies could not corroborate these findings, ^{9 10} the role of estrogen deficiency in the pathogenesis of AMD should be further explored. 

Estrogens are steroid hormones that regulate growth, differentiation, and function of male and female reproductive tracts, mammary glands, and the skeletal and cardiovascular systems. ¹¹ They mediate their effects through two distinct intracellular estrogen receptors (ERs): α(ESR1) and β (ESR2). After binding of estrogen to the ERs, they become transcription factors to modulate gene expression. ¹² The ESR1 gene is located on chromosome 6q25. It consists of at least eight exons and spans more than 400 kb (http://www.ncbi.nlm.nih.gov/SNP/ provided in the public domain by National Center for Biotechnology Information, Bethesda, MD). Many common variations in DNA sequence (polymorphisms) of the ESR1 gene have been identified, including single nucleotide polymorphisms (SNPs). The two most studied SNPs are the adjacent PvuII (rs2234693), a T→C transition in intron 1, and XbaI (rs9340799), a G→A transition located 46 bp downstream of the PvuII polymorphism. The SNPs are localized in the first intron, 397 and 351 bp upstream of exon 2. It is not yet clear whether they have functional consequences. However, these SNPs have been associated with several different phenotypes, such as osteoporosis, Alzheimer’s disease, and cardiovascular disease. ^{13 14 15 16} Recently, it has been demonstrated that, in postmenopausal women, ESR1 PvuII-XbaI haplotype 1 is associated with decreased serum estradiol levels in an allele-dose dependent manner. ¹⁷  

The ESR1 protein is expressed in a variety of tissues, and the presence of ESR1 has also been demonstrated in the retina, suggesting that estrogens may have a role in retinal biology. ^{18 19 20} We set out to study whether SNPs in the ESR1 gene are a risk factor for AMD. 

Methods

Population

The Rotterdam Study is a prospective, population-based cohort study investigating the incidence and determinants of chronic disabling diseases in the elderly. 

All inhabitants aged 55 years or older living in one suburb of Rotterdam, The Netherlands, were invited to participate in the study. ²¹ Of the 10,275 eligible individuals, 7983 (78%) participated. The ophthalmologic part of the study started after screening of the participants had begun, leading to 6780 ophthalmic participants—again, a 78% response rate. The study was conducted according to the tenets of the Declaration of Helsinki, and the medical ethics committee of the Erasmus University approved the study protocol. A written informed consent was obtained from all persons. Baseline examinations, including a home interview and physical examination at the research center took place between March 1990 and July 1993. Three follow-up examinations took place from September 1993 to the end of 1994, from 1997 to 1999, and from 2000 to the end of 2004. 

Definition of AMD

To diagnose AMD, 35° color transparencies were taken of the macular area of each eye (TRV-50VT fundus camera; Topcon Corp., Tokyo, Japan) after dilation of the eyes with tropicamide 0.5% and phenylephrine 5%. These images (digitized from the last follow-up examination to the present) were graded with 12.5× magnification according to the International Classification and Grading System for age-related maculopathy (ARM) by the same two trained professionals who graded AMD from baseline to the present, who were masked for all other determinants. ^{21 22 23 24} We only changed the terminology from early and late ARM in early and late AMD. We categorized the range of AMD fundus signs into five mutually exclusive stages 0 to 4 that had an increasing risk of late AMD. ²⁵ No AMD was defined as stage 0, no signs of AMD at all or only hard drusen (<63 μm); stage 1 was soft distinct drusen (≥63 μm) or only pigmentary abnormalities. Because many participants with only one large druse or one hyperpigmentation in this system are classified as stage 1 and because we wanted to separate participants with marked AMD from those with only limited signs, in the present analyses, we considered stage 1 as no AMD. Thus, we classified as early AMD stage 2, soft, indistinct drusen (≥125 μm) or reticular drusen only, or soft, distinct drusen (≥63 μm) with pigmentary abnormalities, and as stage 3, soft, indistinct drusen (≥125 μm) or reticular drusen with pigmentary abnormalities. Stage 4 was similar to late AMD, subdivided into dry (geographic atrophy) and wet (neovascular) AMD. ²⁵ The disease in each person was classified according to the highest stage of AMD in either eye. Early incident (i)AMD was defined as no AMD at baseline and early AMD in at least one eye at follow-up. Late iAMD was classified as either no or early AMD at baseline and presence of late AMD in either eye at follow-up. Lesions that were considered to be the result of generalized disease, such as diabetic retinopathy, chorioretinitis, high myopia, trauma, congenital diseases, or photocoagulation for reasons other than for wet AMD, were excluded from the AMD diagnosis. 

Genotyping

DNA was extracted from peripheral leukocytes according to standard procedures. All participants were genotyped for the PvuII (rs2234693; c.454-397T→C) and the XbaI (rs9340799; c.454-351A→G) SNPs. Genotypes were also determined for the complement factor H Y402H polymorphism, which is a major risk factor for AMD (described later). Genotypes were determined in 5 ng genomic DNA with the allelic discrimination assay (Taqman; Applied Biosystems, Foster City, CA). Primer and probe sequences were optimized with the SNP assay-by-design service of Applied Biosystems. Reactions were performed on a sequence detection system (Taqman Prism 7900HT; ABI) 384-well format. We used the genotype data for each of the two ESR1 SNPs to infer the haplotypes for each individual using the haplotype reconstruction program PHASE. ²⁶ The alleles were defined as haplotypes, such as T-A representing a thymidine (T) nucleotide at the PvuII polymorphic site and an adenosine (A) nucleotide at the XbaI polymorphic site. We studied haplotypes based on these adjacent polymorphisms, to enhance genetic resolution. Haplotype alleles, based on the combination of these two SNPs, were coded as haplotype numbers 1 through 4 in order of decreasing frequency in the population (1, T-A; 2, C-G; 3, C-A; and 4, T-G). ^{27 28}  

Assessment of Confounders

Information on all potential confounders was collected at baseline. During a home interview, participants were asked by trained research assistants about smoking habits, and women about age at menopause and use of hormone replacement therapy. Smokers were categorized as current, past, or never. In the research center, systolic and diastolic blood pressures were measured twice at the right brachial artery with a random zero sphygmomanometer with the participant in a sitting position. The average of these two measurements was used to determine blood pressure levels. Body mass index was calculated as weight in kilograms divided by height in meters squared. Genotypes of complement factor H (CFH Y402H) polymorphism (1277T→C, rs1061170), were determined in 2-ng genomic DNA extracted according to standard procedures from leukocytes with the allelic discrimination assay (Taqman; ABI). Confounder data for smoking, body mass index, and systolic and diastolic blood pressure were missing in 1% of the participants, and data for CFH genotype in 10%. In the women, confounder data on hormone replacement therapy and age at menopause were missing in 4% of the participants. 

Population for Analysis

At baseline, gradable fundus transparencies were available for 6418 participants, of whom 476 (7.4%) had early AMD and 106 (1.7%) late AMD. Of the 6312 persons at risk for any AMD, 4914 (77.9% of those at risk) participated in at least one follow-up examination and in 4571 (72.4%) of them, haplotype data were available. Persons did not participate in follow-up examinations due to refusal, death, or loss to follow-up. Haplotype data were missing for persons who at baseline only had an interview and did not visit the research center, refused blood sampling, or from whom no blood was available due to various logistic reasons. For each risk analysis for a subtype of AMD, the other AMD cases were removed from the study population as indicated in table footnotes. 

Data Analysis

One-way analysis of variance (ANOVA), for continuous variables, and Pearson’s χ² for dichotomous variables were used to compare possible confounders between participants grouped by the haplotype of interest. The observed genotype distributions were compared using Pearson’s χ² test to determine whether they were in Hardy-Weinberg equilibrium. The associations between ESR1 haplotype stratified on allele copy number (0, 1, or 2) and early or late iAMD, stratified on type of end stage were investigated with the Cox proportional hazards model to compute hazard ratios (HRs) and corresponding 95% confidence intervals (CIs). To calculate a trend, ESR1 haplotype 1 was used as a continuous measure in the model. As a reference group, we used persons with only stage 0 or 1 AMD at baseline and at follow-up. Follow-up time in years was used as time axis of the model. HRs can be interpreted as relative risks. All analyses were adjusted for age and sex. To account for potential confounding, additional adjustments were made for smoking, and the complement factor H Y402H SNP, major risk factors associated with AMD. ^{29 30 31 32} Further adjustment for blood pressure and body mass index did not essentially alter our results (data not shown). Based on previous analyses we chose haplotype 1 as the risk allele. ^{15 28 33 34} To evaluate the effect of sex we performed sex-specific analyses for early and late iAMD. In women, analyses were additionally adjusted for the potential confounding effects of age at menopause and any use of HRT. All analyses were performed with commercial software (SPSS, ver. 11.0; SPSS Inc., Chicago, IL). 

Results

Baseline characteristics of the 4571 participants are shown in Table 1 . ESR1 haplotype 1 was not associated with any of the baseline characteristics, including age at menopause and use at any time of hormone replacement therapy. After an average follow-up time of 7.7 years (range, 0.3–13.9 years), 639 (14.0%) persons had iAMD, of whom 544 (11.9%) had early iAMD, and 95 (2.1%) had late iAMD, of whom 38 had dry and 57 wet iAMD. For early iAMD cases the average follow-up time from baseline to the development of AMD was 4.5 years and for late iAMD cases 6.2 years. We observed the four possible PvuII-XbaI haplotype alleles in the following frequencies: haplotype 1 (T-A) 53.0%, haplotype 2 (C-G) 35.0%, and haplotype 3 (C-A) 12.0%, whereas haplotype 4 (T-G) was not present in our study population. Genotype and allele distributions were in Hardy-Weinberg equilibrium. 

ESR1 haplotype 1 was associated with late iAMD (Table 2) . Persons with one copy of haplotype 1 were at 3.39 (95% CI, 1.62–7.11) times higher risk and for persons with two copies, this was 3.20 (95% CI, 1.47–6.99). After stratification by type of end stage (Table 3) , it seemed that particularly wet iAMD was responsible for this association. Individuals with one copy had a 3.72 (95% CI, 1.31–10.55) times higher risk of wet iAMD, whereas this was 4.71 (95% CI, 1.62–13.66) for individuals with two copies. HRs were also increased for dry iAMD, but they did not reach significance. Early iAMD was not associated with ESR1 haplotype 1. Additional correction for smoking and complement factor H also did not essentially alter our results. After stratification on sex (Table 4) , HRs were increased in the men with one or two copies of haplotype 1, but this only reached statistical significance in the men with one copy and the risk of late iAMD. In the women we found an association with late iAMD: women with one copy of haplotype 1 had a 3.08 (95% CI, 1.07–8.87) times higher risk and this was for the women with two copies 3.62 (95% CI, 1.21–10.77). Further stratification on type of end stage was not possible due to lack of power. Multivariate adjustment, including age at menopause and any use of HRT, did not substantially change these HRs. 

Additional stratified analysis by genotypes of the CFH Y402H polymorphism to test for gene–gene interaction and by smoking status to test for gene–environmental interaction did not show any significant associations (data not shown). 

Discussion

In this study, we demonstrated that the ESR1 PvuII-XbaI haplotype 1 was strongly associated with late iAMD. After stratification by type of end stage, wet iAMD showed significant HRs. Although point estimates were elevated in dry iAMD, there was no significant effect visible, probably due to lack of power. Therefore, it remains inconclusive whether an association with ESR1 PvuII-XbaI haplotype 1 exists. 

The Rotterdam Study is a large study with a population-based prospective design, which increases the reliability of the associations found. Fundus images were graded in a standardized way by the same two well-trained graders at baseline and at all three follow-up visits. Differential misclassification is unlikely, because AMD graders were masked for the presence of ESR1 PvuII-XbaI haplotype 1 status and haplotype data were collected without knowledge of AMD status. There was loss to follow-up due to the older age of the participants, and thus only the healthier persons were able to participate in the follow-up visits. This leads to an underestimation of the strength of the associations. If especially persons with one or two copies of ESR1 haplotype 1 with iAMD had not participated, selection bias would have been introduced. We think this is unlikely because people were unaware of their ESR1 haplotype 1 status, and only in case of late iAMD would they be aware of symptoms. Genetic association studies can be influenced by population heterogeneity. In our study, 99% of the participants were Dutch whites and represent an ethnically homogeneous and representative sample of the population from The Netherlands. 

We could not replicate the association between ESR1 PvuII-XbaI haplotype 1 and late AMD in our prevalent data. The possibility exists that our associations are a chance finding. However, the Rotterdam Study is a large study with a population-based prospective design, which increases the reliability of the associations found. We hypothesized that selective mortality may be involved. Persons with one or two copies of ESR1 haplotype 1 have a higher risk of AMD, and if these persons also die earlier, it could explain the discrepancy between prevalent and iAMD. An independent incidence study is needed to confirm our findings. 

How could these ESR1 SNPs influence the risk of iAMD? A hypothesis regarding functionality of these polymorphisms is that expression of ESR1 is changed through altered binding of transcription factors, perhaps as a direct result of the PvuII and/or XbaI polymorphisms or through linkage disequilibrium with one or more functional sequence variations elsewhere in the ESR1 gene within the linkage disequilibrium block. In support of this hypothesis, it was recently demonstrated that the PvuII C allele produces a functional binding site for the transcription factor B-myb, and the PvuII T allele eliminates this site, suggesting a potential functional polymorphism. ³⁵ The presence of this allele may result in a substantially lower ESR1 transcription or production of isoforms with different features; therefore, if a decreased amount of ESR1 is present or ESR1 with an altered sensitivity for estrogen, estrogen signaling may be less effective. This resembles a situation in which estrogen activity is decreased. ^{15 35} Carriers of the T-allele would be at higher risk for diseases associated with low estrogen levels, such as AMD. 

Recently an association was found between ESR1 PvuII-XbaI haplotype 1 and decreased estradiol levels in a small randomly selected subset of postmenopausal women (n = 631) in the Rotterdam Study. ¹⁷ We could not analyze the association between iAMD and estradiol levels due to low power. It was hypothesized that the lower ESR1 expression caused by the PvuII T allele leads to a lower expression of an enzyme in the estrogen synthesis pathway and thus reduces estrogen levels. ¹⁷ In all men, estrogen serum levels are substantially higher than in postmenopausal women, and this could partially mask the genotypic reduction. This fits with the fact that ESR1 PvuII-XbaI haplotype 1 was the most strongly related with iAMD in the women. Lower serum levels of estrogen in men and especially in postmenopausal women, particularly if they have one or two copies of this haplotype, could lead to reduced transcription of the ESR1. ¹⁸ Whether these ESR1 variants have functional consequences must be investigated, as we lack definite evidence at this moment. In addition, we could not detect an association between HRT and iAMD, probably due to the low number of women taking HRT (n = 69) at baseline. ³⁶  

Several studies have demonstrated the presence of ERs in human, bovine, and rat retinas, principally in the retinal pigment epithelium and choroid. ^{18 19 20} Retinal pigment epithelium ERs are functional, and transcriptional active and estrogen-mediated regulation of genes is important in regulating the turnover of extracellular matrix and in maintaining its basement (Bruch’s) membrane. ¹⁹ Dysregulation in the production of matrix metalloproteinase-2, a gelatinase regulated by estrogens that degrades extracellular matrix components, is thought to contribute to the formation of deposits around Bruch’s membrane. ^{12 37} Because ESR1 haplotype 1 was not associated with early iAMD, we assume that this haplotype influences more the progression of early AMD toward wet AMD instead of the initiation of drusen. Deposits around Bruch’s membrane may promote a proangiogenic response, by binding to integrins or by modifying integrin expression on endothelial cells. This response stimulates the growth of choroidal neovascularization, ³⁸ the basic mechanism of wet AMD. 

In conclusion, we showed an increased risk of wet iAMD in persons with one or two copies of ESR1 PvuII-XbaI haplotype 1. This association was not explained by additional risk factors, marking ESR1 PvuII-XbaI haplotype 1 as a possible independent risk factor for wet iAMD. This study supports a pathophysiological role of estrogen in the development of AMD. 

 Supported by the Netherlands Organization for Scientific Research (NWO), The Hague; the following foundations: Optimix, Amsterdam; Physico Therapeutic Institute, Rotterdam; Neyenburgh, Bunnik; Blindenpenning, Amsterdam; Sint Laurens Institute, Rotterdam; Bevordering van Volkskracht, Rotterdam; Blindenhulp, The Hague; Rotterdamse Blindenbelangen Association, Rotterdam; OOG, The Hague; kfHein, Utrecht; Ooglijders, Rotterdam; Prins Bernhard Cultuurfonds, Amsterdam; Van Leeuwen Van Lignac, Rotterdam; Verhagen, Rotterdam; Netherlands Society for the Prevention of Blindness, Doorn and Elise Mathilde, Maarn; and an unrestricted grant obtained from Topcon Europe BV, Capelle aan de IJssel.

 Submitted for publication May 29, 2006; revised September 19 and October 31, 2006; accepted January 4, 2007.

 Disclosure: S.S. Boekhoorn, None; J.R. Vingerling, None; A.G. Uitterlinden, None; J.B.J. Van Meurs, None; C.M. van Duijn, None; H.A.P. Pols, None; A. Hofman, None; P.T.V.M. de Jong, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Baseline Characteristics of 4571 Participants, Stratified on Number of Allele Copies of ESR1 Haplotype 1 (T-A)

Number of Allele Copies ESR1 Haplotype 1	0	1	2	P
Number of participants (%)	1008 (22.1)	2285 (50.0)	1278 (28.0)
Age (y)	67.1 (7.6)	67.1 (7.6)	66.9 (7.7)	0.7
Sex, female (%)	581 (57.6)	1322 (57.9)	739 (57.8)	0.9
Smoking status (%)
Never	346 (34.5)	760 (33.6)	416 (32.9)
Current	229 (22.8)	496 (21.9)	305 (24.1)	0.5
Past	428 (42.7)	1008 (44.5)	545 (43.0)	0.7
Systolic blood pressure (mm Hg)	137.0 (21.2)	137.7 (21.4)	137.9 (21.8)	0.6
Diastolic blood pressure (mm Hg)	73.7 (11.0)	73.7 (11.0)	74.1 (11.2)	0.5
Body mass index (kg/m2)	26.3 (3.6)	26.3 (3.6)	26.4 (3.7)	0.9
Complement factor H Y402H SNP (%)
Noncarrier	422 (43.6)	939 (42.9)	520 (42.4)
Heterozygous	429 (44.4)	977 (44.6)	556 (45.3)	0.9
Homozygous	116 (12.0)	274 (12.5)	151 (12.3)	0.9
Age at menopause (y)	48.7 (4.9)	48.8 (5.1)	49.1 (4.8)	0.3
Ever use of HRT (%)	84 (15.1)	199 (15.7)	101 (14.2)	0.7

 Data are the means with standard deviations in parentheses, unless otherwise indicated.

Table 2.

 

Risk of Early or Late iAMD According to ESR1 Haplotype 1 (T-A)

Number of Copies of ESR1 Haplotype 1	Participants	Cases (%)	Risk-Adjusted for Age and Sex	Risk-Adjusted for Age, Sex, Smoking, and CFH Genotype
Early AMD
0	1000	116 (11.6%)	1.00 (reference)	1.00 (reference)
1	2228	265 (11.9%)	1.07 (0.86–1.33)	1.04 (0.83–1.30)
2	1248	163 (13.1%)	1.18 (0.93–1.50)	1.16 (0.91–1.48)
P-trend			0.16	0.21
Late AMD
0	892	8 (0.9%)	1.00 (reference)	1.00 (reference)
1	2020	57 (2.8%)	3.39 (1.62–7.11)	3.07 (1.46–6.46)
2	1115	30 (2.7%)	3.20 (1.47–6.99)	2.85 (1.30–6.25)
P-trend			0.01	0.02

 Early AMD: n = 4571 − 95 late AMD cases. Late AMD: n = 4571 − 544 early AMD cases.

Table 3.

 

Risk of Dry or Wet Late iAMD According to ESR1 Haplotype 1 (T-A)

Number of Copies of ESR1 Haplotype 1	Participants	Cases (%)	Risk-Adjusted for Age and Sex	Risk Adjusted for Age, Sex, Smoking, and CFH Genotype
Dry AMD
0	888	4 (0.5%)	1.00 (reference)	1.00 (reference)
1	1989	26 (1.3%)	3.13 (1.09–8.98)	2.76 (0.95–8.01)
2	1093	8 (0.7%)	1.72 (0.52–5.71)	1.41 (0.41–4.85)
P-trend			0.57	0.79
Wet AMD
0	888	4 (0.5%)	1.00 (reference)	1.00 (reference)
1	1994	31 (1.6%)	3.72 (1.31–10.55)	3.50 (1.23–9.94)
2	1107	22 (2.0%)	4.71 (1.62–13.66)	4.29 (1.47–12.49)
P-trend			0.001	0.01

*  Dry AMD: n = 4571 − 544 early AMD cases − 57 wet AMD cases. Wet AMD: n = 4571 − 544 early AMD cases − 38 dry AMD cases.

Table 4.

 

Risk of Early or Late iAMD According to ESR1 Haplotype 1 (T-A) Stratified on Sex

Number of Copies of ESR1 Haplotype 1	Participants	Cases (%)	Risk Adjusted for Age	Risk Adjusted for Age, Smoking, and CFH Genotype*
Men
Early AMD
0	423	53 (12.5%)	1.00 (reference)	1.00 (reference)
1	935	114 (12.2%)	1.01 (0.73–1.40)	1.02 (0.73–1.42)
2	530	74 (14.0%)	1.14 (0.80–1.63)	1.15 (0.80–1.66)
P-trend			0.43	0.42
Late AMD
0	374	4 (1.1%)	1.00 (reference)	1.00 (reference)
1	849	28 (3.3%)	3.22 (1.13–9.19)	3.02 (1.05–8.72)
2	465	9 (1.9%)	1.88 (0.58–6.09)	1.71 (0.51–5.71)
P-trend			0.49	0.58
Women
Early AMD
0	577	63 (10.9%)	1.00 (reference)	1.00 (reference)
1	1293	151 (11.7%)	1.12 (0.83–1.50)	1.10 (0.81–1.49)
2	718	89 (12.4%)	1.21 (0.88–1.68)	1.13 (0.81–1.61)
P-trend			0.24	0.48
Late AMD
0	518	4 (0.8%)	1.00 (reference)	1.00 (reference)
1	1171	29 (2.5%)	3.51 (1.23–9.99)	3.10 (1.07–8.90)
2	650	21 (3.2%)	4.55 (1.56–13.26)	3.52 (1.18–10.46)
P-trend			0.01	0.03

*  Plus age at menopause and use (ever) of HRT in women.

The authors thank Ada Hooghart and Corina Brussee, who graded the AMD images, and Pascal Arp, who genotyped the ESR1 SNPs.