Article Date: 9/1/2013

Combining Genetics With Disease Stage To Optimize AMD Management

Combining Genetics With Disease Stage To Optimize AMD Management

Factoring genetic predisposition into risk assessment leads to improved detection of CNV and reduced monitoring burden.

Gregory Hannum, PhD, is a biostatistician at Sequenom, Inc., in San Diego. Lorah T. Perlee, PhD, is the former vice president of scientific affairs at Sequenom. Both authors report significant financial interest in the company. Dr. Perlee can be reached via e-mail at


Genetic testing for risk stratification in patients with age-related macular degeneration opens the door to improving the efficiency and effectiveness of current management protocols.

The American Academy of Ophthalmology task force on genetic testing issued guidelines in December 2012, that recommend physicians:

… avoid routine genetic testing for genetically complex disorders like age-related macular degeneration and late-onset primary open-angle glaucoma until specific treatment or surveillance strategies have been shown in one or more published clinical trials to be of benefit to individuals with specific disease associated genotypes.1

Currently, clinicians rely on clinical features to predict the likelihood a patient will progress from the earlier stages of AMD to the advanced forms, choroidal neovascularization or geographic atrophy.2

While accurate clinical staging can provide a strong basis for assessing the risk of disease progression, patients with similar stages of disease progress to CNV at different rates.


A recent study by Perlee et al3 aimed to capture the clinical significance of combining a patient’s genetic profile with his or her stage of disease to evaluate the impact on CNV prediction accuracy. The study demonstrated improved accuracy of predicting progression to CNV, based on a retrospective analysis of clinical data derived from 2,415 subjects participating in the Age-Related Eye Disease Study.3

The AREDS trial was one of the largest multicenter, prospective, longitudinal studies the National Eye Institute/National Institutes of Health conducted.4 The study evaluated the clinical course of AMD and cataract over 10 years and was well-powered for developing and validating genetic risk prediction methodologies to enhance clinical care.6

Table 1. Frequency of Patient Surveillance In Various Patient Groups by Stage of Disease and Risk

Prescribed Surveillance in Weeks to Achieve <2.5% CNV Conversion Per Risk Group RetnaGene AMD Low Risk (Bottom 25% of CNV risk profiles for a given grade) RetnaGene AMD Moderate Risk (Middle 50% of CNV risk profiles for a given grade) RetnaGene AMD High Risk (Top 25% of CNV risk profiles for a given grade)
Grade 0 (39%) 1074 wks (21 yr) 693 wks (13 yr) 438 wks (8.4 yr)
Grade 1 (17%) 365 wks (7.0 yr) 213 wks (4.1 yr) 120 wks (2.3 yr)
Grade 2 (16%) 156 wks (3.0 yr) 83 wks (1.6 yr) 41 wks (9.4 mo)
Grade 3 (12%) 78 wks (1.5 yr) 36 wks (8.3 mo) 15 wks (3.4 mo)
Grade 4 (15%) 52 wks (1.0 yr) 20 wks (4.6 mo) 10 wks (2.3 mo)

The findings of the Perlee et al study revealed that genetic factors could account for variations in the rate of choroidal neovascular conversion in patients with the same baseline disease stage, resulting in greater accuracy in predicting progression.

The ability to use multiple risk factor analysis to stratify patient populations has significant clinical and practical implications. Although CNV impacts less than 10% of the AMD disease population, it accounts for greater than 90% of the vision loss associated with AMD.6

While no clinical intervention is currently available to reduce progression to GA, treatment regimens using anti-VEGF therapies have proved effective in reducing the near-term CNV-associated vision loss in some patients. Importantly, timely diagnosis remains central to minimizing the irreversible vision-loss associated with treatment delay.7-9

We expand on this earlier work to translate the utility of risk stratification into benefits gained in the clinical management of dry AMD patients.



The AAO preferred practice pattern guidelines include monitoring early or intermediate-stage dry AMD patients every six to 24 months. This broad recommendation reflects the wide variation in the timing of presentation of CNV, prediction of which can only succeed with the use of clinical classifications, such as the AREDS AMD grading system.

The Perlee et al3 study demonstrated that genetic information builds upon the predictive value of phenotypic disease grade in predicting the likelihood and timing of CNV development.

In comparison, estimates of disease progression, based on clinical phenotypic information alone, likely result in clinicians overestimating the risk and monitoring frequency in patients with low-risk genotypes, and underestimating the risk and monitoring frequency in those with high-risk genetic profiles.

Perlee et al3 established the strong correlation between high-risk genetic profiles and CNV progression, suggesting the clinical assessment of a patient’s genotype may positively influence screening intervals and other methods of surveillance, providing a sound rationale for future prospective studies.

We are not aware of any prior publications that have combined genetic predisposition to AMD with patient disease stage to provide the scientific basis for establishing risk-stratified monitoring protocols.

In this study, we demonstrate how we can use patient risk assessment data, based on genetic predisposition, disease stage, age, and smoking status, to model improved patient surveillance regimens.

Modeling performed in a group of 2,415 AREDS subjects, previously reported in Perlee et al,3 showed clinicians can achieve potential gains in disease management to reduce the number of conversion events before the first follow-up visit following genetic testing without increasing the overall monitoring burden. The findings reveal the benefits of applying an individualized approach to AMD patient management.


White, non-Hispanic subjects participating in the AREDS trial consented to provide a genetic specimen. Of 2,415 DNA specimens previously analyzed for progression to CNV,3 940 were from disease-free subjects, and 1,475 were from subjects with early or intermediate AMD at baseline.

We used the AREDS simplified severity scale (SSS)10 assignment at baseline to identify a subgroup of phenotypically high-risk subjects, to evaluate the association of overall CNV risk profile and surveillance frequency in this study.

This high-risk subgroup consisted of 655 subjects exhibiting one of the following four characteristics:

• drusen in both eyes and pigmentary change in one eye (AREDS SSS grade 3, bilateral dry);

• drusen or pigmentary change in one eye and advanced disease in the fellow eye (AREDS SSS grade 3, unilateral advanced);

• both drusen and pigmentary changes in both eyes (AREDS SSS grade 4, bilateral dry); or

• drusen and pigmentary changes in one eye and advanced disease (CNV and/or GA) in the fellow eye (AREDS SSS grade 4, unilateral advanced).

Table 2. Fixed-interval Schedules and Optimized Risk-based Schedules

Prescribed Frequency of Visitation Using Fixed Interval Monitoring vs Risk-Based Monitoring For Grade 3, 4 Subjects RetnaGene AMD Low Risk (Bottom 25% of CNV risk profiles for a given grade) RetnaGene AMD Moderate Risk (Middle 50% of CNV risk profiles for a given grade) RetnaGene AMD High Risk Top 25% of CNV risk profiles for a given grade)
24 months Fixed-interval monitoring 104 wks 104 wks 104 wks
  Risk-based monitoring grade 3 158 wks 117 wks 86 wks
  Risk-based monitoring grade 4 143 wks 97 wks 75 wks
18 months Fixed-interval monitoring 78 wks 78 wks 78 wks
  Risk-based monitoring grade 3 122 wks 88 wks 64 wks
  Risk-based monitoring grade 4 109 wks 73 wks 55 wks
12 months Fixed-interval monitoring 52 wks 52 wks 52 wks
  Risk-based monitoring grade 3 83 wks 59 wks 43 wks
  Risk-based monitoring grade 4 74 wks 49 wks 36 wks
9 months Fixed-interval monitoring 39 wks 39 wks 39 wks
  Risk-based monitoring grade 3 63 wks 45 wks 32 wks
  Risk-based monitoring grade 4 56 wks 36 wks 27 wks
6 months Fixed-interval monitoring 26 wks 26 wks 26 wks
  Risk-based monitoring grade 3 44 wks 30 wks 21 wks
  Risk-based monitoring grade 4 37 wks 24 wks 18 wks


We developed a risk-stratification approach to assign optimal patient surveillance frequency based on the outcome of the RetnaGene AMD Laboratory Developed Test (LDT; Sequenom, Inc., San Diego, CA). We designed the surveillance regimen to reduce the number of conversions that occurred before the first scheduled follow-up visit, while maintaining the total number of patient exam visits relative to a fixed monitoring schedule.

We classified the subjects according to their grade of disease (eg, AREDS simplified severity scale grade 4) and their risk assignment, based on the RetnaGene AMD output. We developed the algorithm by evaluating the RetnaGene AMD risk profiles for the 2,415 AREDS subjects relative to the known CNV outcomes at interim visits every six months.

We defined high risk as a risk profile that was in the top 25% of CNV risk profiles observed in the AREDS study population for all subjects with the same stage of disease,3 while moderate and low risk reflected the middle 50% and bottom 25% of CNV risk profiles, respectively.

To ensure robust and generalizable estimates, we based the optimized monitoring schedules and associated statistics on risk models generated from the 2,415 AREDS subjects, with randomized repeated validation based on 500 simulated trials. Each trial consisted of a training set chosen from two-thirds of the data and a test set consisting of the remaining third.

To simulate population variances, we resampled each training set with replacement (bootstrapping). For each trial, we constructed a model using only the training set and the backward regression approach used in Perlee et al.3 The reported statistics for optimized monitoring times and performance are based on the median of the results from each of the trained models on its respective test set.

Main Outcome Measurements

Choroidal neovascular conversion rates and visitation frequencies were evaluated for the fixed interval regimens and compared with the surveillance prescribed using the RetnaGene AMD risk stratification approach to identify the best method for surveillance.


Progression Risk and Surveillance Frequency

One straightforward way to assign surveillance frequencies is to model the risk for each patient population and match her or his visitation frequency directly to her or his risk. We used this approach to assign a frequency of surveillance based on RetnaGene AMD risk profiles, reflecting a patient’s stage of disease, genetic predisposition, age, and smoking history.

Table 1 (page 35) shows the results of the prescribed frequency of surveillance (in weeks) based on five categories of disease stage (AREDS simplified severity scale 0-4) and three categories of risk (low, moderate, and high), generating 15 RetnaGene AMD risk groups.

We designed the algorithm to achieve a conversion rate of less than 2.5% before the first follow-up visit for each of the 15 risk groups. Based on these results, the specified monitoring time can differ more than fivefold between patients with the same stage of disease.

Matching risk of conversion before the first follow-up visit (eg, <2.5%) within each risk group (eg, Grade 4 RetnaGene high) is only one of several ways to optimize surveillance schedules. Following such a schedule does not fully account for the overall visitation burden placed on the physicians and patients.

Optimization Algorithm

To overcome this limitation, we designed an algorithm that optimizes visitation schedules to minimize the number of conversion events before the first follow-up exam, while simultaneously comparing the total number of visits to a fixed exam schedule (such as every six months).

This provides an approach to incorporate genetic information into common clinical practice without changing the overall monitoring burden. Given the low CNV conversion rates observed in patients with no/early stage disease (AREDS simplified severity scale grade 0-2), we applied the algorithm to a targeted group of high-risk patients assigned AREDS simplified severity scale grade 3 or 4.

Table 3. Conversion Rates and Surveillance Burden For AREDS Subjects With Severity of 3 or 4

Performance of Fixed- Interval vs Risk-Based Monitoring Schedules Risk-Based vs Fixed-Interval (24 mo) Risk-Based vs Fixed-Interval (18 mo) Risk-Based vs Fixed- Interval (12 mo) Risk-Based vs Fixed-Interval (9 mo) Risk-Based vs Fixed-Interval (6 mo)
Visitation burden 0.5 visits per year 0.67 visits per year 1 visit per year 1.33 visits per year 2 visits per year
Fixed-interval monitoring conversions 10% (7.1 – 14)% 7.3% (4.7 – 11)% 4.6% (2.7 – 7.1)% 3.4% (1.9 – 5.3)% 2.2% (1.1 – 3.5)%
Risk-based monitoring conversions 9.6% (6.5 – 13)% 6.8% (4.2 – 9.9)% 4.3% (2.4 – 6.5)% 3.1% (1.7 – 4.9)% 2.0% (1.0 – 3.2)%
Conversion rate improvement 7.1% (2.5 – 13)% 7.5% (2.8 – 13)% 7.8% (2.9 – 14)% 8.1% (3.2 – 14)% 8.2% (2.9 – 14)%
Number of conversions per 1000 93 vs 100 68 vs 73 42 vs 46 31 vs 34 20 vs. 22

Table 2 (page 36) shows the optimized schedules that we calculated for five fixed-monitoring schedules at different times (six, nine, 12, 18, and 24 months). This table highlights the difference in monitoring patients based on individualized risk.

As an example, all patients monitored under a six-month fixed protocol would be seen every 26 weeks regardless of risk, whereas using a risk-stratified approach, high-risk Grade 4 patients would be monitored more frequently (every 18 weeks) and low-risk grade 4 patients would be monitored less frequently (every 37 weeks).

Evaluating Conversion Rates Before Follow-up

When compared to fixed-interval monitoring, a risk stratification approach, based on the assignment of a high-, moderate-, or low-risk test outcomes, reduced the rate of conversion (patients that converted to CNV before the first follow-up eye exam) without increasing the number of visits for the patient group.

Table 3 shows the conversion rates and performance of the fixed-monitoring approach vs RetnaGene AMD risk-based monitoring across the high-risk group comprised of AREDS Grade 3 and 4 subjects. The conversion rate improvement achieved using the RetnaGene AMD risk-based monitoring approach was statistically significant when compared to the fixed monitoring regimen for all five visit schedules evaluated (six, nine, 12, 18, and 24 months) (P < .05).

A separate analysis was conducted (results not shown) to determine whether the RetnaGene AMD risk-based monitoring approach demonstrated a conversion rate improvement over a disease stage-only approach (based on the assessment of a patient’s AREDS simplified severity scale and status of CNV in the fellow eye) to evaluate the contribution of genetics.

The results using the RetnaGene AMD risk-based monitoring revealed a conversion rate improvement that was statistically significant compared to a monitoring schedule based on disease staging alone (P < .05).


This study revealed how combining genetic predisposition with patient disease stage could translate risk stratification into more effective patient monitoring regimens without increasing the burden of surveillance in a busy practice.

The modeling results conducted in 2,415 AREDS subjects confirmed the relationship between a patient’s risk of progressing from dry AMD to wet AMD and the ability to align more closely surveillance with time to CNV conversion.

When bench marked to fixed interval regimens (eg, every six, nine, 12, 18, and 24 months), the increased monitoring requirements for high-risk patients contributed to the gains in higher detection rates and were balanced by the decreased monitoring needed in low-risk patients.

These results show how using a risk-stratified approach more closely aligns exams with conversions in high-risk patients and minimizes unnecessary visits in low-risk patients with the same stage of disease.

Additional gains in detection beyond those observed in this study might be achieved by stratifying patients into more granular categories than those defined with RetnaGene AMD. Future studies will investigate these categories further. RP


1. Stone EM, Aldave AJ, Drack AV, et al. Recommendations for genetic testing of inherited eye diseases report of the American Academy of Ophthalmology task force on genetic testing. Ophthalmology. 2012;119:2408-2410.

2. Klein ML, Francis PJ, Ferris FL III, et al. Risk assessment model for development of advanced age-related macular degeneration. Arch Ophthalmol. 2011;129:1543-1550.

3. Perlee LT, Bansal AT, Gehrs K, et al. Inclusion of genotype with fundus phenotype improves accuracy of predicting choroidal neovascularization and geographic atrophy. Ophthalmology. 2013 Mar 20. [Epub ahead of print]

4. Age-Related Eye Disease Study Research Group. The Age-Related Eye Disease Study (AREDS): design implications. AREDS report no. 1. Control Clin Trials. 1999;20:573-600.

5. Chew EY, clemons TE, Agróri E, et al; Age-Related Eye Disease Study Group. Long-term effects of vitamins C and E, images-Carotene, and xinc on age-related macular degeneration: AREDS Report No. 35. Ophthalmology. 2013;120:1604-1611.

6. AMD Alliance International. Basic facts about AMD. Available at: http://www. Accessed February 4, 2013.

7. Lim JH, Wickremasinghe SS, Xie J, et al. Delay to treatment and visual outcomes in patients treated with anti-vascular endothelial growth factor for age-related macular degeneration. Am J Ophthalmol. 2012;153:678-686.

8. Muether PS, Hoerster R, Hermann MM, et al. Long-term effects of ranibizumab treatment delay in neovascular age-related macular degeneration. Graefes Arch Clin Exp Ophthalmol. 2013;251:453-458.

9. Muether PS, Hermann MM, Koch K, Fauser S. Delay between medical indication to anti-VEGF treatment in age-related macular degeneration can result in a loss of visual acuity. Graefes Arch Clin Exp Ophthalmol. 2011; 249:633-637.

10. Age-Related Eye Disease Study Research Group. A simplified severity scale for age-related macular degeneration: AREDS report no. 18. Arch Ophthalmol. 2005;123:1570-1574.

Retinal Physician, Volume: 10 , Issue: September 2013, page(s): 34 - 39