A, Age at onset for each individual based on the number of risk alleles in CFH and ARMS2. B, Effect of smoking on the mean age at onset for each genetic risk stratum.
Lechanteur YTE, van de Camp PL, Smailhodzic D, van de Ven JPH, Buitendijk GHS, Klaver CCW, Groenewoud JMM, den Hollander AI, Hoyng CB, Klevering BJ. Association of Smoking and CFH and ARMS2 Risk Variants With Younger Age at Onset of Neovascular Age-Related Macular Degeneration. JAMA Ophthalmol. 2015;133(5):533–541. doi:10.1001/jamaophthalmol.2015.18
The age at which the first signs of age-related macular degeneration (AMD) manifest is variable. Better insight into factors that influence disease onset has direct implications for preventive measures and patient counseling.
To identify risk factors for an earlier age at onset of neovascular AMD.
Design, Setting, and Participants
Retrospective cohort study, including patient data from the European Genetic Database collected between April 2006 and July 2010. All patients had at least 1 documented visit to the outpatient AMD clinic of the Radboud University Medical Center, Nijmegen, the Netherlands, a tertiary referral center for retinal disorders. In total, 275 patients with a known age at onset of neovascular AMD and a genetic risk analysis were included.
Main Outcomes and Measures
Effects of several genetic, sociodemographic, behavioral, and ocular factors on the age at onset of neovascular AMD. The mean differences in the age at onset were determined using general linear models with the age at onset as the dependent variable.
Past smokers and current smokers developed neovascular AMD on average 4.9 (95% CI, 3.0-6.8) and 7.7 (95% CI, 5.3-10.0) years earlier, respectively, than never smokers (P < .001 for both). Compared with the reference group, the age at onset was 5.2 (95% CI, 2.8-7.7) years earlier for homozygous carriers of the A69S risk allele in the age-related maculopathy susceptibility 2 (ARMS2) gene (P < .001). Homozygous carriers of the Y402H risk variant in the complement factor H (CFH) gene developed neovascular AMD 2.8 (95% CI, 0.5-5.0) years earlier (P = .02). Patients carrying 4 risk alleles in CFH and ARMS2 developed neovascular AMD 12.2 (95% CI, 6.2-18.3) years earlier than patients with zero risk alleles (P < .001).
Conclusions and Relevance
Genetic and environmental risk factors influence the age at onset of neovascular AMD. Individuals at risk could be identified at an early age if and when preventive or therapeutic options become available. Insight into individual risk profiles might influence patients’ consideration of interventions to increase their chance of avoiding vision loss from AMD.
Age-related macular degeneration (AMD) is the most common cause of severe visual impairment among elderly individuals in developed countries. Typically, 2 forms of end-stage disease that can be distinguished are geographic atrophy (GA) (or dry AMD) and neovascular AMD characterized by choroidal neovascularization (CNV). Although only 10% of patients experience neovascular AMD, it accounts for most blindness due to AMD.1,2
Smoking is the strongest modifiable risk factor for AMD.3- 12 Other inconsistent environmental and demographic risk factors include obesity,4,13- 17 sunlight exposure,18- 20 and sex.4,21,22 A healthy diet containing omega-3 fatty acids, lutein, zeaxanthin, and antioxidant vitamins C and E may be protective against the development of AMD.23- 26 Nutritional supplements containing the Age-Related Eye Disease Study 2 formula27 can further reduce risk of progression to advanced AMD in patients at risk. Apart from environmental factors, several genetic variants have been associated with AMD. The strongest associations have been found for single-nucleotide polymorphisms (SNPs) in the complement factor H (CFH) gene (OMIM 134370)28- 30 and in the age-related macular susceptibility 2 (ARMS2) gene (OMIM 611313).31- 33 Besides these variants, other SNPs have been associated with AMD, including variants in genes involved in the complement pathway,34- 37 the high-density lipoprotein cholesterol pathway,38,39 and the atherosclerosis signaling pathway.40,41
To date, much research has focused on the identification of risk factors for the development and worsening of AMD. In contrast, few studies have been conducted to investigate the age at onset of AMD. The age at which the first signs of AMD manifest is variable, and it is important to gain insight into contributing factors. A few studies have reported an influence of genetic and environmental risk factors on the age at onset of AMD. Smoking, male sex, triglyceride levels, and variants in the CFH, ARMS2, GSTM1, and VEGFA genes have been implicated to be associated with a younger age at onset.42- 48 Statin and aspirin use seemed to postpone the time to CNV development in one study,42 but another study48 showed an earlier onset of AMD for statin users. Differences in study results and the magnitude of the effects on the age at onset may arise because of different definitions of the age at onset, inclusion of patients with nonadvanced stages of AMD, and few study participants in some of these studies.
The objectives of this study were to identify additional risk factors that may influence the age at onset and to determine whether conventional risk factors for AMD are also involved in an earlier age at onset of the disease. To define the age at onset, we limited study inclusion to patients with neovascular AMD. This approach enabled us to identify the type and influence of genetic and environmental risk factors important in accelerating the development of neovascular AMD.
This study was conducted in accord with the tenets of the Declaration of Helsinki and was approved by the local ethics committee (Commissie Mensgebonden Onderzoek Regio Arnhem–Nijmegen). All members of the study population provided written informed consent and participated between April 2006 and July 2010. We included patients who had neovascular AMD with at least 1 documented visit to the outpatient AMD clinic of the Radboud University Medical Center, Nijmegen, the Netherlands, and who were genotyped for at least 1 of the SNPs studied. In total, 385 eligible patients were selected from the European Genetic Database (EUGENDA; http://www.eugenda.org), a multicenter database for clinical and molecular analysis of AMD.
The age at onset of neovascular AMD was determined by reviewing historical patient data. Age at onset was defined as when a patient experienced the first visual complaints related to CNV occurrence in either eye, with a diagnosis of CNV by an ophthalmologist within the next 6 months.49 Clinical diagnosis by an ophthalmologist was by slitlamp examination or by evaluation of fundus photography or fluorescein angiography. In case of scarring or other signs indicating that CNV had already been present for a longer time, the patient was excluded from the analysis.
Participants were asked to complete a questionnaire on lifestyle and other environmental variables (Table 1). For classification of smoking habits, patients were asked whether and how long they had smoked. For past smokers, we also obtained the duration of smoking cessation in years. These data were adjusted to match the status at the time of CNV development. For example, a patient who had quit smoking 7 years before the interview and had developed CNV 10 years before the interview was classified as a current smoker. Body mass index was calculated from self-reported weight and height. For sunlight exposure, participants were asked how much time they had spent outside before their retirement. Exposure was classified as low for patients who indicated avoidance of the sun or spending most of the time indoors, moderate for patients who were regularly outside but no longer than 8 hours a day, and high for patients who were outside for more than 8 hours a day. If available, data on refractive error were recorded from the patients’ medical records.
Venous blood samples were collected for extraction of DNA. The DNA was analyzed for SNPs in the following 9 genes previously associated with AMD: CFH, ARMS2, CFB, C3, CFI, APOE, LPL, CETP, and ABCA1 (Table 2).30,33- 40 The SNPs were genotyped using competitive allele-specific polymerase chain reaction assays (KASPar SNP Genotyping System; KBiosciences).
All variables were analyzed using 1-way analysis of variance (ANOVA) without post hoc tests. For some variables, different categories were merged to create subgroups of sufficient size for statistical analysis. Variables with P < .15 were selected for inclusion in a general linear model with the age at onset as the dependent variable. For this model, multiple imputation of missing data values was used. Variables with P ≥ .05 were removed from the model in a stepwise fashion. For the final variables included in the model, parameter estimates were calculated to derive the mean differences in the age at onset for each risk category. All analyses were performed using statistical software (SPSS, version 20.0; IBM Corporation).
Multiple imputation of missing data values was performed under the assumption that observations were missing at random. Only the 7 variables that were included in the multivariable model were imputed. For imputation of each variable, we used the 6 other variables, the outcomes, and the unselected variables from the 1-way ANOVA as predictors. Imputation was performed using the default settings in SPSS (version 20.0). From Tables 1 and 2, rates of missing data for each variable imputed can be calculated. In total, 69.1% of patients had complete data for all 7 variables. Only 9.1% of patients had missing data for more than 1 of the 7 variables.
Data from the Rotterdam Study I (RS-I), a population-based study with more than 20 years of follow-up, were used to plot survival curves for the risk factors that were identified in the multivariable model to estimate the effect of possible survival bias on our results. Details on the RS-I have been described previously.50
Of 385 eligible patients, 275 were included in the study. Complete information on the age at onset was available for 214 participants. For 38 patients, there was no exact information on the duration of visual complaints; in this group, the age at onset was defined as when CNV was diagnosed by an ophthalmologist. In a small subgroup (n = 23), we could only retrieve the year of diagnosis by an ophthalmologist. For this group, July 1 of that year was considered the date of onset. One hundred ten patients were excluded from the study because of incomplete medical records (n = 66), other macular pathology that could interfere with the diagnosis of AMD (n = 36), and treatment in the macular region for reasons other than AMD-related CNV (n = 8).
The mean (SD) age at onset of CNV was 74.8 (7.7) years (range, 53.1-90.7 years), and 56.7% of participants were female. The patients with 1 or more missing variables did not differ significantly from the patients with complete data with respect to the age at onset. Sex, smoking, alcohol consumption, and exercise level were associated with the age at onset of neovascular AMD in the 1-way ANOVA (Table 1). Of the genetic factors analyzed, rs10490924 (ARMS2 A69S), rs1061170 (CFH Y402H), and rs3764261 (CETP) were selected for inclusion in the multivariable model (Table 2).
Of the 7 variables that were selected from the 1-way ANOVA, only 3 remained associated with the age at onset in the multivariable model (Table 3). Compared with never smokers, past smokers developed CNV on average 4.9 (95% CI, 3.0-6.8) years earlier (P < .001). For current smokers, this difference was even larger: they developed CNV 7.7 (95% CI, 5.3-10.0) years earlier (P < .001). Homozygous carriers of the Y402H risk allele in CFH had an earlier disease onset of 2.8 (95% CI, 0.5-5.0) years compared with the reference group (P = .02). For the ARMS2 A69S variant, homozygous carriers developed CNV 5.2 (95% CI, 2.8-7.7) years earlier (P < .001).
To explore the combined effect of CFH and ARMS2 on the age at onset, we delineated patient groups based on the number of risk alleles of these 2 variants. There was a difference in the age at onset of 12.2 (95% CI, 6.2-18.3) years when we compared individuals having zero risk alleles in CFH and ARMS2 with those carrying all 4 risk alleles (P < .001) (Table 4 and Figure, A). Panel B in the Figure shows the substantial effect of smoking on the age at onset in addition to the genetic risk. Current smokers with zero risk alleles developed CNV 14.5 years earlier than never smokers with the same genetic risk profile (P = .08). The sole effect of smoking resulted in an age at onset comparable to that of never smokers carrying all 4 risk alleles. When we compared the most extreme groups in panel B in the Figure, there was an observed absolute difference in the age at onset of 22.4 years (P = .03). Some of these subgroups were small (range, 2-40 patients).
The exact age at onset was sometimes difficult to determine. To exclude the possibility of introducing bias, we reran the analyses in a subgroup, selecting only those patients for whom the age at onset was most accurately defined and diagnosed (duration of visual complaints <6 months, exact date of diagnosis known, and diagnosis confirmed in the Radboud University Medical Center). These results did not differ significantly; to maintain statistical power, we chose to adhere to the analyses of the complete data set.
Analyses with data from the RS-I revealed that there was no difference in survival for the different genotypes in CFH and ARMS2. Survival curves for smoking showed that current smokers had a shorter survival than past smokers and never smokers. Before reaching age 70 years, 7.9% of smokers vs 3.2% of never smokers had died. In total, 79.5% of smokers reached age 75 years compared with 90.2% of never smokers.
The development of CNV accounts for most legal blindness in patients with AMD.1,2 However, the range in the age at onset of CNV is broad, with some patients developing neovascularization as early as age 50 years, whereas others are well past age 90 years before this occurs. On average, the patients in this study were aged 74.8 years when their first CNV developed. Many potential risk factors that might influence the age at onset of neovascular complications in AMD were examined. In the presence of homozygous CFH Y402H and ARMS2 A69S risk alleles, the age at onset was advanced by 12.2 years. In addition, smoking was associated with the development of neovascular AMD at a younger age. When we added smoking to the genetic risk factors, the difference in the age at onset between the most extreme groups increased to 22.4 years. Although these results should be interpreted with caution because of small group sizes in the combination models, no other study has reported effect sizes of this magnitude, to our knowledge. The substantial effect of smoking, CFH, and ARMS2 on the age at onset of CNV has consequences for patient counseling. Based on our results, we hypothesize that many disease-free years may be attained if a patient stops smoking, and it seems likely that other preventive measures (eg, the intake of dietary supplements) could be of benefit as well. Because the development of CNV may occur well before age 70 years in high-risk patients, these countermeasures have to be initiated in a timely manner to be of significance.
The risk factors that remained in the multivariable model probably accelerate the age at onset by increasing complement activity and oxidative stress levels in the retina. Decreased functionality of the CFH protein, for instance in persons with the CFH Y402H polymorphism, leads to a continuous state of low-level complement activation.51,52 Complement activation is known to lead to increased vascular endothelial growth factor (VEGF) expression and CNV formation.52- 54 For ARMS2, some investigators suggest that it is a mitochondrial protein involved in oxidative stress responses,31,55 although this supposition was disputed by others.56 One study57 found that the protein interacts with proteins from the extracellular matrix, while another study58 demonstrated that ARMS2 is involved in the phagocytosis of photoreceptor outer segments. In 2012, it was reported that the ARMS2 risk allele is independently associated with increased complement activation in patients with AMD.59 Cigarette smoke adds to oxidative stress60 and is known to induce VEGF expression as well as retinal pigment epithelium (RPE) dysfunction and diminished RPE cell survival that may affect angiogenic homeostasis.61,62 Moreover, cigarette smoke increases the activity of the alternative complement pathway by weakening the susceptibility of C3 to CFH and factor I.51,63 Smoking, CFH Y402H, and ARMS2 A69S have been associated with CNV development in large independent cohorts.3,4,44,64- 66 We hypothesize that increased oxidative stress, a continuous state of complement activation, and upregulation of VEGF expression not only increase risk of CNV development but also may accelerate the age at which neovascularization occurs.
The association of CFH variants with early onset of neovascular AMD was suggested in a 2011 study,45 in which a rare penetrant mutation in CFH was associated with a 6-year earlier onset of disease in patients with advanced AMD. Recently, another study48 confirmed the association of CFH and ARMS2 with an earlier onset of AMD and showed that heavy smokers are also affected with AMD at a younger age. In that study, the largest difference in the age at onset reported was only 2.2 years for heavy smokers vs the reference group, while for CFH and ARMS2 there was less than a 2-year difference in onset between homozygous carriers of the risk variants compared with those carrying the low-risk variants. These differences are small compared with our findings. However, it is unclear how the age at onset was defined in the study by Keilhauer et al,48 and their study included patients with CNV and patients with early AMD. Inclusion of different AMD stages makes it impossible to uniformly determine the age at onset in all participants, and this could be reflected in the study outcomes. Several other studies42,46,67 evaluating the age at onset restricted inclusion to patients with CNV, as described above. However, these study populations were small (≤131 patients with CNV), and racial/ethnic differences hinder comparison with the present study.
This study identified much larger effect sizes of CFH, ARMS2, and smoking on the age at onset than previously reported by including only individuals with neovascular AMD and using a clear definition of the onset of neovascular complications. Because of the nature of the EUGENDA database, we were able to study the effect of multiple genetic and environmental risk factors on the age at onset of neovascular AMD, whereas other studies42,46,48,67 limited their design to few variables. This protocol enabled us to also study the combined effect of genetic and environmental risk factors.
However, a retrospective study design has limitations. A possible confounding factor is the subjective nature of the duration of visual complaints and the decision as to whether these complaints are in fact related to CNV development. However, subgroup analyses that included only those patients with a well-defined age at onset yielded results similar to those of the main analyses.
Another possible confounder is the interference of survival bias. Survival curves from the RS-I showed that 79.5% of smokers reached age 75 years (the mean age at onset in this study) compared with 90.2% of never smokers. This result makes it unlikely that survival bias accounts for the effect of smoking on the age at onset that we reported. Furthermore, smoking can be seen as an accelerator of aging directly (through oxidative stress) and indirectly (through smoking-associated diseases).68,69 In this light, it may be that smokers have the same chance of developing neovascular AMD as never smokers, although their survival is shorter.
The present study investigated potential risk factors for the age at CNV onset and did not include patients with GA. Because there is a strong overlap in risk factors for CNV and GA development,65,66,70- 72 we hypothesize that the same risk factors that contribute to the development of GA also lead to an earlier age at onset of GA. It would be difficult to analyze this in a retrospective study because the exact onset of GA in patients with AMD is difficult to define.
In individuals who are homozygous for CFH and ARMS2 risk alleles and who smoke cigarettes, the onset of neovascular AMD may be accelerated by as much as 1 or 2 decades. To allow time for countermeasures to take effect, these individuals need to be identified at an early age. Providing individuals with a personalized prognosis may give them the incentive to stop smoking. Severe visual loss associated with the development of neovascular membranes could be postponed by additional preventive measures, such as a healthy diet, use of nutritional supplements, frequent monitoring, and possible future therapeutic options aimed at lowering complement activity.
Submitted for Publication: July 28, 2014; final revision received December 29, 2014; accepted December 30, 2014.
Corresponding Author: B. Jeroen Klevering, MD, PhD, Department of Ophthalmology, Radboud University Medical Center, PO Box 9101, 6500 HB Nijmegen, the Netherlands (email@example.com).
Published Online: February 19, 2015. doi:10.1001/jamaophthalmol.2015.18.
Author Contributions: Drs Lechanteur and Klevering had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hoyng, Klevering.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Lechanteur, van de Camp.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Lechanteur, Buitendijk, Groenewoud.
Obtained funding: den Hollander, Hoyng.
Study supervision: Klevering.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest, and none were reported.
Funding/Support: This study was supported by MD Fonds, Oogfonds, and Algemene Nederlandse Vereniging ter Voorkoming van Blindheid.
Role of the Funder/Sponsor: The supporting organizations had no role in any of the following: design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.