Abstract
Purpose:
To determine behavioral and genetic factors associated with incidence and age of progression to advanced age-related macular degeneration (AMD), geographic atrophy (GA), and neovascular disease (NV), and to quantify these effects.
Methods:
Longitudinal analyses were conducted among 5421 eyes with nonadvanced AMD at baseline in 2976 participants in the Age-Related Eye Disease Study (mean age of 68.8 (±5.0), 56.1% female). Progression was confirmed based on two consecutive visits on the AMD severity scale. Separate analyses for progression and age of progression were performed. All analyses adjusted for correlation between eyes, demographic and behavioral covariates, baseline severity scale, and genetic variants.
Results:
A higher genetic risk score (GRS) including eight genetic variants was associated with a higher rate of progression to advanced AMD within each baseline severity scale, especially for the highest risk intermediate level AMD category, and smoking further increased this risk. When assessing age when progression to advanced disease occurred, smoking reduced age of onset by 3.9 years (P < 0.001), and higher body mass index (BMI) led to earlier onset by 1.7 years (P = 0.003), with similar results for GA and NV. Genetic variants associated with earlier age of progression were CFH R1201C (4.3 years), C3 K155Q (2.15 years), and ARMS2/HTRA1 (0.8 years per allele).
Conclusions:
Rare variants in the complement pathway and a common risk allele in ARMS2/HTRA1, smoking, and higher BMI can lead to as much as 11.5 additional years of disease and treatment burden. Closer adherence to healthy lifestyles could reduce years of visual impairment.
Age-related macular degeneration (AMD) has a complex cause and remains a significant public health problem despite recent advances in treatments.
1–3 Many patients with neovascular macular degeneration (NV) disease have residual visual impairment after treatment with intravitreal injections, because of varying degrees of chorioretinal atrophy and scarring. Therefore new therapies are being evaluated. The advanced dry form with geographic macular atrophy (GA) has no known treatment, but many clinical trials are underway.
AMD confers a significant individual and societal burden and can lead to loss of independence, increased use of health care resources, and an adverse impact on quality of life. The prevalence of AMD is increasing as the proportion of our elderly population rises, and the number of people with AMD is expected to be 196 million in 2020, increasing to 288 million in 2040.
2 AMD is the leading cause of visual disability in the developed world and the third globally.
4 The estimated direct health care costs of visual impairment in North America caused by AMD is $98 billion and $255 billion globally.
5 The level of reduction in quality of life for severe AMD is comparable to end-stage prostatic cancer or a catastrophic stroke.
6 Prevention of AMD is therefore a key public health strategy.
A set of genetic, demographic, and environmental variables can predict with relatively high likelihood which individuals will more likely progress to advanced AMD.
7 It has also been shown that individuals with rare genetic variants are more likely to progress.
7–9 However, the age at which the transition from nonadvanced to advanced AMD occurs is variable, even among those with the same baseline macular pathology. The independent effect of individual genetic variants and behavioral variables on this age of progression, and quantification of the difference in number of years, have not been evaluated in a longitudinal study. We analyzed data from a large, well-defined cohort to assess the impact of both genetic and lifestyle factors on age when transition to advanced AMD occurs over time, adjusting for other known factors related to AMD.
Analyses included some new methodologic considerations. AMD incidence is not a linear function of age; risk increases nonlinearly after age 70.
2,7,10 Thus we used age as the time scale in conducting time-to-event analyses. Severity scales for each eye were analyzed separately to better adjust for baseline macular status. Also, to enhance accuracy in determining the endpoints, we defined a progressing eye on the basis of having a severity scale indicating advanced disease at two consecutive visits. Also, because the effect of the association between severity scale and AMD is not linear, we represented severity scale as a set of indicator variables in the analyses. Finally, we used stepwise regression to identify relevant or predictive genotypes in deriving a genetic risk score on the basis of genotypes related to specific outcomes in longitudinal analyses.
Clinical trials should consider selection criteria that target particular disease subgroups for therapeutic approaches, such as those at higher risk of progression or have earlier age of progression with longer disease burden. To help achieve this goal, the aims of these analyses were to (i) apply new methods for evaluating predictors of developing overall advanced AMD, GA, and NV, and (ii) evaluate the impact of both modifiable and genetic risk factors on age when progression occurs. Herein, we expand upon our previous preliminary analyses and results on this topic.
11,12
The Genetic Risk Score (GRS
AdvancedAMD) was calculated using the variants associated with progression to overall advanced AMD (
Table 1). Based on this model, we calculated a GRS = sum of β
i, multiplied by g
i , where g
i = number of risk alleles present for the ith genetic variant i=1,…,8:
\begin{equation*}GRS = \mathop \sum \limits_{i = 1}^8 {\beta _i}{g_i}\end{equation*}
Table 1. Genetic Variants Associated With Progression to Overall Advanced AMD, GA, and NV
Table 1. Genetic Variants Associated With Progression to Overall Advanced AMD, GA, and NV
The GRS was then grouped into tertiles, the severity scale was categorized into four groups (1–5/6/7/8) and a second Cox model was run using age as the time metric included GRS tertile and severity scale category and in addition controlling for sex, education, race, BMI, smoking, and AREDS treatment group. We used the baseline statement of SAS with the second survival model to estimate the survival curve for a subject with average value for all covariates.
In the third step of the analysis, we used the baseline survival curve to estimate the five-year survival probability (S) with specific combinations of severity scale and GRS tertile, and average levels of all other covariates:
\begin{equation*}S\left( 5 \right) = \;{\left[ {{S_{baseline}}\left( 5 \right)} \right]^{{e^{\left( {\mathop \sum \nolimits_{i = 2}^3 {\beta _i}GR{S_i} + \;\mathop \sum \nolimits_{j = 2}^4 {\gamma _j}SE{V_j}} \right)}}}}\end{equation*}
where GRS
i and SEV
j for the ith GRS and jth severity scale category, respectively.
We then estimated the five-year incidence, I(5) = 1-S(5), for each combination of GRS category and severity scale category. Similar GRSs were also constructed for GA (GRSGA) and NV (GRSNV). Similar plots were constructed for different GRS tertiles within combinations of smoking status (current/past/never) and representative severity scales (2/5/8), for 12 years.
Genetic Risk Score and AUC Analyses for Progression to Advanced Age-Related Macular Degeneration
Table 2 displays the GRS for progression to each outcome derived from the selected genetic variants in
Table 1. These scores were significantly associated with all three outcomes, adjusting for all covariates including the baseline severity scale (
P value <0.001 for all outcomes). For advanced AMD overall, the highest tertile of GRS was associated with a threefold increase in rate of progression compared to the lowest tertile, and almost a threefold increased risk of progression between the ninetieth versus tenth percentile. For GA, the risks were somewhat less pronounced between highest and lowest tertiles than for overall advanced AMD (HR of 1.90), with more than a twofold higher rate of progression for the ninetieth versus tenth percentile. For NV, the risk was 3.7-fold higher for the third GRS tertile compared to the first tertile, and similarly was 3.6 fold higher for ninetieth versus tenth percentile.
Table 2. GRS for Progression to Advanced AMD, GA, and NV
Table 2. GRS for Progression to Advanced AMD, GA, and NV
Figure 1 displays Kaplan-Meier (KM) survival curves for progression according to GRS tertiles, adjusting for all covariates, using the baseline command of PROC PHREG of SAS. The estimated 12-year risk of progression was 12%, 20%, and 25% for tertiles 1, 2, 3; indicating about a twofold increased risk for GRS tertile 3 versus 1.
Figure 2 displays KM curves for GRS tertiles stratified by severity scales and smoking status. There is little difference in survival according to GRS for severity scale of 2 (early signs of AMD) and a small proportion of subjects progressed by 12 years. Conversely, there were large differences in survival by GRS categories for severity scale 5 and especially for scale 8, which were magnified further among individuals who were current smokers. For example, among past smokers with baseline severity scale of 8, the 12-year survival probability was 45% for GRS 1 versus 10% for GRS 3. For the current smokers with a severity scale of 8, the 12-year survival probability was 20% for GRS 1 versus less than 5% for GRS 3. Similar findings were apparent for severity scale 5. In summary, GRS differences in survival were only apparent for subjects with at least intermediate AMD, and most discriminating among subjects with the later preadvanced stages of intermediate disease.
Figure 3 displays the 5 year and 12 year cumulative incidence of progression to advanced AMD, GA, and NV, for various combinations of baseline severity scale and GRS, which adjust for competing mortality risks. For overall progression, there was a strong effect of severity scale on incidence of AMD. Within specific levels of the baseline severity scale, there was an additional gradient of risk when subdividing by tertiles of GRS, particularly in severity scale range 6–8. For example, for severity scale 8, there was approximately a 50% cumulative incidence of advanced AMD over 12 years if the GRS was in the lowest tertile compared with 95% cumulative incidence if the GRS was in the highest tertile. However, if the severity scale was in the lower range 1 to 5, there was a low probability of progression at five or 12 years, regardless of the GRS.
Subjects in the highest tertile of GRS and the higher severity scales (6–8) had an increased risk for progression to GA when mutually adjusting for GRS and AMD baseline scale. The incidence rates for progression to NV in eyes with baseline scales 6 to 8 were similar, although within each of these baseline grades, similar to GA, the third tertile of GRS had the highest risk of progression, followed by tertile 2, and the lowest risk was seen in the lowest GRS for each baseline scale 6, 7, and 8.
As shown in
Table 3, the AUC for the five-year incidence of advanced AMD was 0.88 among the group with baseline severity scale 1 to 8, adjusting for only demographic and behavioral factors, and was 0.89 after adding the GRS, indicating that there is only a small increase in discrimination when adding the GRS (
P = 0.032). Because this AUC calculation is dominated by the large number of eyes in severity scale range 1 to 5, a group with low risk of progression, we also restricted the analysis to eyes within baseline scales 6–8. The AUC was lower but was substantially improved by including tertiles of GRS in addition to the severity scale for overall AMD (ΔAUC = 0.039,
P < 0.001). This is illustrated by at least a threefold difference in incidence rate of progression comparing GRS tertile 3 to tertile 1 for individuals within severity scale 6 to 8 (
Fig. 3). The effect of the GRS was also more pronounced when restricting eyes to higher risk baseline severity scales 6 to 8 for progression to NV (Δ = 0.062,
P < 0.001). In addition, for NV there appears to be a threshold effect of severity scale with a large difference in risk of progression between severity scales 1 to 5 versus 6 to 8, but little change in risk among severity scales 6 to 8. Overall, these results indicate that the GRS provides additional differential information, particularly for eyes at high risk of progression (with baseline scales 6–8).
Table 3. AUC for Progression to Advanced AMD, GA, and NV Over Five Years According to Baseline Severity Scale, With and Without Genetic Variables
Table 3. AUC for Progression to Advanced AMD, GA, and NV Over Five Years According to Baseline Severity Scale, With and Without Genetic Variables
Table 4 displays the multivariate analysis of the effects of demographic, behavioral, ocular, and genetic factors on
age of progression to advanced AMD, GA, and NV. The average age among progressors is shown for a reference group with none of the risk factors: female, nonwhite, higher education, never smoker, normal BMI, baseline severity scale 1, and none of the genetic risk variants.
Table 4. Multivariate Analysis of Associations Between Demographic, Behavioral, Ocular, and Genetic Factors and Age of Progression to Advanced AMD, GA, and NV Among Eyes that Progressed
Table 4. Multivariate Analysis of Associations Between Demographic, Behavioral, Ocular, and Genetic Factors and Age of Progression to Advanced AMD, GA, and NV Among Eyes that Progressed
Subjects who were current smokers (P < 0.001) or had higher BMI ≥30 (P = 0.003) had an earlier age at progression to advanced AMD relative to the average age ( 3.9 years and 1.7 years, respectively), compared with the reference categories of never smoking or having a BMI < 25. Similarly for GA, current smoking and higher BMI were associated with earlier age of progression (3.5 years and 2 years, respectively). Smoking was associated with 3.24 years earlier progression to NV.
Baseline severity scale 8 compared to scale 1 was associated with earlier age of progression to AMD (average of 3.2 years earlier age of progression; P = 0.007). There was a similar trend for GA (average of 4.4 years earlier, P = 0.073). A significant trend for earlier age of progression for increasing severity scale was seen for both advanced AMD and GA.
Three variants were found to be associated with age of progression to advanced AMD: CFH R1210C: rs121913059 with an average of 4.3 years earlier age at progression among carriers of this mutation compared with non-carriers (P = 0.026), C3 K155Q: rs147859257 with an average of 2.15 years earlier age at progression for carriers (P = 0.027), and ARMS2/HTRA1 A69S rs10490924 with an average of 0.79 years earlier age of progression per risk allele (P = 0.006). For carriers of the homozygous genotype for ARMS2/HTRA1, with two risk alleles, the impact would be an average of 1.58 (2 × 0.79) years earlier age of progression. For GA, only CFH R1210C was associated with earlier age of progression (average of 5.4 years earlier among carriers, P = 0.033). For NV, only ARMS2 was found to be associated with earlier age of progression (average of 0.93 years per risk allele, P = 0.018).
Figure 4 shows the distribution of age of progression to advanced AMD among progressors according to various genetic and non-genetic subgroups.
Figures 4A to
4C display the mean age of progression shifted toward younger ages for the common
ARMS2/HTRA1 risk allele and the rare
CFH and
C3 variants. In the homozygous
ARMS2/HTRA1 risk group, approximately 45% of eyes progressed at age <75, compared with 30% for the homozygous nonrisk group. In addition, for the rare variant
CFH R1210C, approximately 65% of carriers vs 30% of non-carriers had age of progression <75. A similar but weaker trend was seen for carriers vs non-carriers of the
C3 K155Q variant. Approximately 50% of current smokers vs 30% of never smokers had an age of progression <75. Similar trends were seen when considering combinations of risk factors (see
Figures 4E,
4F). Approximately 50% to 55% of subjects who were current smokers and also homozygous
ARMS2 carriers had age of progression <75, compared with 30% of never smokers and
ARMS2 nonrisk genotype. Combinations of current smoking and high BMI showed a similar trend of earlier age of progression.
Supplementary Figure S3 shows the box plots of the median and distribution of age of progression among subjects grouped by smoking status and BMI status, with the category of current smokers plus highest BMI showing the earliest age of progression.
Table 5 displays the association between GRS
age and
age of progression to advanced AMD incorporating three genes related to this outcome. Age of progression was reduced by approximately 1.9 years among subjects with two or more risk alleles; furthermore, age of progression was reduced by 2.5 years per one unit increase in GRS.
Table 5. Association Between GRS Incorporating CFH R1210C, C3 K155Q, and ARMS2, and Age of Progression to Overall Advanced AMD
Table 5. Association Between GRS Incorporating CFH R1210C, C3 K155Q, and ARMS2, and Age of Progression to Overall Advanced AMD
Tables 6A and
6B show population-attributable risks for modifiable behavioral and genetic factors. For never and past smokers, the percentage of progression to advanced AMD prevented by optimizing their BMI to <25 was 11.6% and 12.5%, respectively. Conversely, for current smokers, optimizing their behavior (i.e., changing to past smoker and BMI to <25) would prevent approximately 60% of progression to advanced AMD. Over for the entire study population, 14.7% of disease could be prevented by optimizing modifiable behaviors. Similarly, approximately 37% of disease progression would be prevented if GRS profile changed from second and third tertile to the first tertile; however, this is not modifiable.
Table 6A. Population Attributable Fraction for Behavioral Factors
Table 6A. Population Attributable Fraction for Behavioral Factors
Table 6B. Population Attributable Fraction for GRS Factors
* Table 6B. Population Attributable Fraction for GRS Factors
*
We determined that behavioral factors modify risk of progression to advanced AMD, GA, and NV, and genetic variation at multiple AMD risk alleles is associated with increased rate of progression after adjusting for baseline severity scale and all covariates. In other words, for each level of baseline severity scale, especially for the higher risk scales 6–8, time to conversion to late AMD decreased with increasing genetic risk and with unhealthy behaviors.
Age when progression to advanced AMD occurred was also affected by genes and environment. Smoking and higher BMI were associated with earlier age of progression and thus more years of disease and treatment burden. Genetic susceptibility also played a role, and higher genetic burden due to carrying high risk rare variants in CFH or C3 in the complement pathway, or the common risk allele in the ARMS2/HTRA1 gene, were associated with earlier age of transitioning from non-advanced to advanced AMD. Smoking together with a higher BMI were associated with 5.6 years earlier age of progression to advanced disease, and genetic burden determined by a common variant, e.g. ARMS2/HTRA1, was associated with lower age of progression by 1.6 years for homozygous carriers (combined total of up to 7.2 years). If an additional rare variant, such as CFH R1210C, was present, then disease onset would be shortened by an additional 4.3 years, or as much as 11.5 years in total earlier onset of advanced disease and longer disease burden, adjusting for all other covariates.
The greatest impact of genetic factors on enhancing prognostic ability to predict who will progress to advanced AMD, especially for NV, was seen among eyes with baseline scales 6–8, which represent increasing levels of intermediate disease with larger drusen. This is important because these are the eyes at highest risk for developing advanced stages causing visual loss. It appears therefore that there are three distinct subgroups: eyes with severity scale 1–5 with low risk regardless of their genetic burden, eyes with intermediate risk, which have severity scales 6–8 but a low genetic burden, and another group of eyes with high risk, which have severity scales of 6 to 8 and a high genetic burden. For eyes with severity scales 6 to 8 and high genetic burden (highest tertile of GRS), about 35% will progress to advanced AMD over five years, whereas about 10% of eyes with baseline scales 6 to 8 with low genetic risk (lowest tertile of GRS) and <5% of eyes with baseline scales of 1 to 5 developed advanced AMD over 5 years.
In summary, in this first longitudinal analysis of the impact of both genetic and non-genetic factors on age of progression to GA and NV, results underscore the impact of nature and nurture on both developing advanced disease leading to visual loss, and also increasing the likelihood of having this adverse outcome at an earlier age. A combined risk of smoking, higher BMI and genetic factors could lower age of developing advanced stages of AMD by 7 to 11.5 years, leading to a longer burden of disease and more treatments. Thus, adhering to healthy habits and changing unhealthy habits could reduce societal and economic costs.
Knowledge of factors underlying age of transition to advanced AMD can impact clinical care by underscoring the importance of adhering to healthy habits and provides biologic mechanisms to explore to shorten disease burden. Accounting for individual differences in lifestyles, as well as genetic susceptibility, may lead to avenues for personalized medicine. We will ultimately be able to target different disease subgroups in clinical trials and offer more precise therapeutic approaches.
Supported by NIH R01-EY011309, R01-EY028602, R01-EY022445, American Macular Degeneration Foundation, Northampton, MA, The Macular Degeneration Center of Excellence, University of Massachusetts Medical School, Department of Ophthalmology and Visual Sciences, Worcester, MA.
Disclosure: J.M. Seddon, Scientific Co-Founder, Gemini Therapeutics, Inc.; R. Widjajahakim, None; B. Rosner, None