Does Genetic Variation Influence Response to Treatment with Vascular Endothelial Growth Factor Inhibitors in Neovascular Age-related Macular Degeneration?

Over the past decade, treatments based on inhibiting the activity of vascular endothelial growth factor (VEGF) have transformed the care of patients with neovascular age-related macular degeneration (AMD). The most widely used anti-VEGF agents include ranibizumab, bevacizumab and aflibercept, and it is noteworthy that the functional outcomes from all three drugs are highly effective. However, despite this remarkable clinical effect, a wide range in treatment response has been observed.1,2 As genetic variation has been shown to be a significant risk factor in the development of neovascular AMD, attention has now focused on the potential influence of these genetic risk alleles in determining responsiveness to treatment by anti-VEGF biologicals. The loci harboring genetic markers that account for the majority of the known population-attributable risk, complement factor H (CFH) and age-related maculopathy susceptibility 2 (ARMS2) genes, have been subjected to pharmacogenetic analysis in a variety of studies with differing results. In a few small-scale case series, it has been shown that the presence of CFH and ARMS2 risk alleles are associated with adverse visual outcomes.3 However, no statistically significant pharmacogenetic associations were observed between these polymorphisms and visual acuity outcomes or anatomical outcomes with data collected from large, randomized clinical trial settings.4,5

While many explanations likely exist for this lack of association between genetic risk and response to anti-VEGF treatment, the first important reason is the bias-free nature of using material derived from large clinical trials with prospectively defined eligibility criteria and phenotypic characterization of neovascular AMD with robust outcome metrics. Second is the selection of the most appropriate outcome of responsiveness to treatment. Visual acuity is a function of retinal morphology at the fovea and does not completely reflect amelioration or exacerbation of pathology in other regions of the retina involved in neovascular AMD such as the parafovea and the majority of the macula. The third possible reason is the absence of biological plausibility. It should be questioned why genes in the immunological pathway, for example, that initiate neovascular growth could be involved in modulating the response to anti-VEGF treatments which merely mute the exudative manifestations. And fourth, as the effects of the drugs are short-lived, repeated administration is not often achieved in routine clinical settings. Therefore, treatment outcomes may be suboptimal thus confounding the detection of true pharmacogenetic associations.

In a related article by van Asten and colleagues in this issue of Ophthalmic Epidemiology,6 researchers from the Netherlands present data using a modified approach to the analysis of association between specific genetic loci and treatment with anti-VEGF agents. They report that predictors for non-responsiveness, defined as a loss of visual acuity by 30 letters from treatment initiation measured at 3 months, was influenced by baseline acuity, diabetes mellitus and accumulation of risk alleles in CFH, ARMS2 and VEGF-A genes. Strengths of this study include (1) a more robust analysis of potential predictors using a receiver operator curve method which adjusts for the various potential confounders, and (2) the use of an outcome metric of visual acuity loss at 3 months at which point in time variability in re-treatment decision-making is avoided as all patients receive monthly treatment in this type of loading phase. Limitations of this study include (1) small to moderate sample size and resultant statistical power, (2) dependence upon Snellen visual acuity as opposed to Early Treatment Diabetic Retinopathy Study (ETDRS) vision charts for the majority of patients, and (3) patients were recruited through routine clinical visits. However, in our opinion, the most fundamental limitation in this study is the use of a functional metric as a surrogate for responsiveness rather than a direct reflection of treatment effect such as a morphological outcome. Despite complete disappearance of exudative manifestations in eyes with neovascular AMD treated with anti-VEGF therapy, losses in visual acuity can occur as a consequence of structural changes. For example, retinal and/or retinal pigment epithelial (RPE) atrophy, fibrosis involving the outer retina and/or RPE layer and tears of the RPE are all incompatible with good visual function even though the exudative manifestations from neovascular AMD have been controlled. Therefore, we would argue that the most appropriate definition for treatment responsiveness would be a metric that captures the amelioration of the exudative response.

Neither of the two large, randomized clinical trials (CATT and IVAN) that analyzed pharmacogenetic associations found evidence of a role for genetic variation in determining responsiveness to treatment when based on either functional or morphological outcomes.4,5 Furthermore these trials also found no association between treatment response and genetic variation in the genes encoding VEGF or its cognate receptor where there is more biological plausibility.5,7

Unless robust evidence on pharmacogenetic associations are forthcoming, we contend that genetic testing as a step towards predicting clinical response is not warranted and that at present there is no rationale for modifying therapy for individuals based on their genetic profiles.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.