Genetic Testing for AMD Offers High Sensitivity and Specificity

Available tests can enhance the clinical examination.

CARL C. AWH, MD

Genetic tests for AMD have been commercially available for more than five years, but events during the past year have brought increased attention to this tremendous technology.

In December 2012, the American Academy of Ophthalmology’s Task Force on Genetic Testing issued the following Guidelines for Genetic Testing.

• Ensure that the patient receives counseling from a physician with expertise in inherited diseases.

• Use Clinical Laboratories Improvement Amendment–approved (CLIA) laboratories for all genetic testing.

• Give patients a copy of test reports for use in seeking additional information and potential clinical trials.

• Discourage use of direct-to-consumer genetic testing.

• Order the most specific tests available to suit the patient’s clinical findings.

• Avoid routine genetic testing for genetically complex disorders, like AMD and late-onset primary open-angle glaucoma, until specific treatment or surveillance strategies have been shown.

• Avoid testing asymptomatic minors for untreatable disorders, except in extraordinary circumstances.

The AAO Task Force also stated:

Genetic testing can make a very positive impact on individuals and families affected with inherited eye disease in a number of ways. When properly performed, interpreted, and acted upon, genetic tests can improve the accuracy of diagnoses and prognoses, improve the accuracy of genetic counseling, reduce the risk of disease occurrence or recurrence in families at risk, and facilitate the development and delivery of mechanism-specific care.

The AAO Task Force has provided excellent guidance for physicians. Moreover, the latest generation of genetic tests for AMD completely satisfies the criteria established by the AAO.

In this article, I’ll provide information about one such test, the MaculaRisk NXG test from Arctic DX. However, my comments are also generally relevant for another test, RetnaGene from Sequenom.

BACKGROUND

As recently as seven years ago, there was little understanding of the specific genetic basis for AMD. Newer and more affordable methods of genetic analysis and research and data from tens of thousands of study subjects have contributed to a more complete understanding of the genes associated with AMD, increasing our understanding of the pathogenesis of the disease and paving the way for earlier identification of patients at risk.

AMD CANDIDATE GENES

Approximately 12 million sites of historical mutation along the human genome have been catalogued and measured. These sites of variation are known as single nucleotide polymorphisms (SNPs). SNPs are sites of base pair substitution (eg, adenine-thymine substituted for cytosine-guanine). SNPs do not necessarily have any functional effect; rather, they serve as markers for the adjacent segment of DNA.

Studies have been performed to find associations between commonly occurring SNPs and a multitude of diseases. Of all human multigenetic diseases, AMD stands out as having the strongest genetic contribution, with approximately 12 SNPs accounting for the majority of the disease risk.¹

The initial AMD “candidate genes” were discovered through genome-wide association studies (GWAS), in which the DNA of individuals with advanced AMD was compared with the DNA of persons within a control group. Through this process, SNPs associated with AMD were identified.

Figure 1. Relative impact of genetic risk markers on advanced AMD.

Figure 2. The progression to different stages of AMD is driven by different genes.

IMAGE FROM YU Y, REYNOLDS R, ROSNER B, DALY MJ, SEDDON JM. PROSPECTIVE ASSESSMENT OF GENETIC EFFECTS ON PROGRESSION TO DIFFERENT STAGES OF AGE-RELATED MACULAR DEGENERATION USING MULTISTATE MARKOV MODELS. INVEST OPHTHALMOL VIS SCI. 2012;53:1548-1556. REPRODUCED WITH PERMISSION FROM THE ASSOCIATION FOR RESEARCH IN VISION AND OPHTHALMOLOGY.

Figure 3. Probability curves for true- and false-positive and -negative results.

The SNPs associated with AMD are found on genes that influence pathways consistent with our understanding of the biology of the disease. For example, several identified SNPs are linked with genes that play important roles in the complement cascade.

The complement cascade is considered to be part of the “innate” immune system, and it is known to contribute to inflammation.

Components of the alternative complement pathway implicated in the pathogenesis of AMD include complement factor H (CFH), complement component 2 (C2), complement factor B (CFB), complement component 3 (C3), and complement factor I (CFI).¹

The core of AMD initiation may lie in an imbalance of components within the complement cascade. Remarkably, independent predictors of disease development and progression have been discovered and confirmed within multiple components of a single physiological process. This level of understanding has not been duplicated in any other major human multigenic disease.

Other genes associated with AMD appear to be important for generating oxygen free radicals. The most strongly associated gene is the age-related maculopathy 2 gene — ARMS2. ARMS2 is likely involved in energy metabolism, inferred from its subcellular localization to mitochondrial outer membranes.

CALCULATING GENETIC RISK

Genetic risk marker identification has provided an unprecedented ability to calculate the lifetime genetic risk of developing AMD. Because each risk marker acts independently, risk in an individual having more than one marker can be estimated by the summed probability of the individual risk markers. Given the number and magnitude of known risk markers, the resultant summed risk score can vary tremendously.

For example, the ratio of risk of an individual with two unfavorable copies of the CFH, ARMS2, and C3 genes to the risk of an individual without any of these risk markers is almost 400 (odds ratio).

The degree to which genetics contributes to the risk of AMD has not been demonstrated in any other common human disease of aging (including diabetes, heart disease, hypertension, or stroke).

The relative contribution of these and other genetic variables to AMD risk can be represented by the pie chart in Figure 1. So-called “minor genes” account for more than half of the calculated risk for late-stage AMD. If these genes could be eliminated from the human gene pool, the prevalence of AMD might be cut in half.

The first generation of AMD genetic tests used the known odds ratios of risk markers to calculate a “lifetime” genetic risk, based only upon measured SNPs. This “risk score” was valuable mainly when integrated into the clinician’s interpretation of nongenetic factors (clinical AMD status, age, etc.) for each patient.

More recently, research in AMD genetics has advanced from identifying predictors of lifetime risk to fully integrated analyses that promise to predict the transition from one stage of AMD to another.

MULTISTAGE MODELS FOR PREDICTING DISEASE PROGRESSION

AMD is a disease characterized by a long period of early disease with a transition to intermediate disease and finally to one of two late-stage disease states. Having a large set of disease-associated markers makes it possible to determine whether all markers are equally important in influencing each transition or whether they act differentially.

In a study recently published in Investigative Ophthalmology and Visual Science, Yu and coworkers¹ used multistate Markov modeling, a method of probability assessment, to assess prospectively the effects of genetic markers on progression within the different stages of AMD.

A typical patient with AMD transitions from normal (stage 1) to intermediate-size drusen (stage 2); then from intermediate to large drusen (stage 3); and then from large drusen to advanced AMD (stage 4), manifested as geographic atrophy or choroidal neovascularization.

Some genetic drivers of each transition are common, but some may be unique to particular transitions and might be independent of the risk factors that delivered an individual patient to his or her current risk state.

In Yu and coworkers’ Markov model of AMD disease progression,¹ four types of transitions are defined, and the epidemiological and genetic drivers can be separated. Previous studies have indicated the importance of all of these markers for the ultimate development of latestage disease. This new approach to understanding disease transition provides some interesting insights.

For instance, an elevated body-mass index is significant for the transition from intermediate AMD (AREDS 3) to GA, but not for the development of CNV. Smoking affects transition to GA and CNV but does not appear to be significant for the development of AREDS 3 disease. Age is an important driver of CNV development but does not appear to be a significant influence in the progression of AREDS 3 to GA.

Figure 4. A visual explanation of sensitivity and specificity, using the area under a receiver operating characteristic curve.

Figure 5. The ROC AUC for five-year progression is 0.883, while the ROC AUC for 10-year progression is 0.895 in Yu and Seddon’s model of AMD progression.

Similar to epidemiological and demographic variables, genetic variables appear to act differently at each disease transition state. For instance, the two major determinants of AMD — CFH and ARMS2 — act to drive the transition from AREDS 2 to AREDS 3 disease, as well as AREDS 3 to CNV or to GA.

Figure 6. The curves from Figure 5, with the optimized point for sensitivity/specficity trade-off findicated.

In contrast CFB, LIPC and CETP appear to be the most significant in driving the transition from AREDS 3 to CNV, while CFI is a significant cant driver of progression from AREDS 3 to GA. Gene variants, such as ABCA1 and COL8A1, appear to drive the development of AREDS 3 disease, but they do not have a measurable influence on the development of CNV or GA. In this model C2 and APOE variants do not have any measurable effects.

Taking the statistical observations, a genetic pathophysiological model of disease progression can be constructed (Figure 2, page 56). Normal individuals are predisposed to the development of intermediate drusen if they have risk variables in the ABCA1, CFI, and LIPC genes.

The transition to large drusen develops under the additional influence of CFH and ARMS2, as well as C3 and COL8Al. The transition from a state of large drusen to CNV is favored by risk variables in the LIPC, CETP, and CFB genes, while progression to GA occurs under the influence of CFI. Background genes, such as C3, CFI, and ARMS2, drive progression in general.

EVALUATING TESTS FOR AMD PROGRESSION RISK

The addition of next-generation AMD risk markers allows for the unprecedented ability to determine individual risk of progression to advanced AMD. Understanding these markers may also allow us to identify patients at increased risk for specific transitions from one stage of AMD to another (eg, progression to GA vs CNV).

Meaningful prediction of individual risk for progression to advanced-stage AMD will facilitate the identification of individuals at increased risk. These patients might be the best candidates for early intervention before the loss of vision.

MEANINGFUL PREDICTION OF RISK

What represents a “meaningful” prediction of risk? First, the four relationships between the test outcome and reality must be defined. A test can be positive with disease present (true positive: TP), a test can be negative with disease absent (true negative: TN), but the test can also provide false information (Figure 3, page 56). It can be positive when no disease exists (false positive: FP), and it can be negative when, in fact, the disease is present (false negative: FN).

The sensitivity of a test is the ability of a test to detect disease when it is present (TP/all with disease), and the specificity of a test is the ability to correctly identify those that do not have disease (TN/all without disease). By shifting the criteria for “positive,” the sensitivity and specificity of a test can be altered.

For example, a test that always generates a positive result will always identify disease (perfect sensitivity) and will never produce false negatives, but it would have terrible specificity and produce many false positives. An example of this would be a pregnancy test that always tests positive (even in men!).

The genetic and nongenetic risk factors in an individual produce essentially a continuous variable as the test output. For each “cut-off” to define “positive,” sensitivity and specificity can be determined and plotted (sensitivity vs 1-specificity), generating the ROC curve.

The area under the curve (AUC) is a measure of the accuracy of the prediction parameters (Figure 4, page 57). An AUC of 0.5 is no better than random chance (a coin toss), and an AUC of 1.0 would be a perfect predictive model (perfect sensitivity and specificity).

Most medical screening tests have an AUC of 0.7 or better. The Framingham coronary heart disease risk score (FCRS) has an AUC of 0.63 to 0.83, depending upon the population studied. The AUC for Yu and Seddon’s predictive model of AMD progression is 0.895 (Figure 5, page 57), demonstrating that the variables measured seem quite capable of predicting advanced AMD. This curve also suggests a natural threshold for positivity (shown in blue) that would result in greater than 80% sensitivity and specificity (Figure 6).

SUMMARY

Tremendous progress has been made in the field of AMD genetics. Early discoveries (CFH and ARMS2) have stood the test of time and are important indicators of the overall risk of developing late-stage AMD in those that have early AMD.

Next-generation techniques and marker discoveries have added tools that may help predict an individual’s risk of progressing to GA or to CNV, or an unaffected person’s risk of developing early AMD. Together, these marker sets allow for unprecedented accuracy in the prediction of AMD outcome with near perfect accuracy.

One goal of genetic testing is to identify individuals who may progress to advanced AMD so that actions can be taken to influence their outcomes favorably. At present, such actions remain limited, as there are no reliable preventive therapies for advanced AMD.

However, the results of genetic testing can be used to increase patient awareness and compliance and to identify patients for whom more intensive surveillance is reasonable. These simple steps should lead to earlier detection of treatable advanced AMD, as well as a significant overall reduction of vision loss in at-risk patients. RP

REFERENCES

1. Yu Y, Reynolds R, Rosner B, Daly MJ, Seddon JM. Prospective assessment of genetic effects on progression to different stages of age-related macular degeneration using multistate Markov models. Invest Ophthalmol Vis Sci. 2012;53:1548-1556.

2. Li M, Atmaca-Sonmez P, Ohtham M, et al. CFH haplotypes without the Y402H coding variant show strong association with susceptibility to age-related macular degeneration. Nat Genet. 2006;38:1049-1054.