Subjects

A retrospective chart review was conducted for 220 patients from a single institution (New England Eye Center, Boston, Massachusetts, USA) who had been concurrently or previously enrolled in genetic and epidemiologic studies of the etiology of AMD, and who were diagnosed with neovascular AMD. This research adhered to the tenets of the Declaration of Helsinki and was performed under appropriate institutional review board protocol from the New England Eye Center and Ophthalmic Epidemiology and Genetics Service at Tufts Medical Center. Signed informed consent was obtained from all subjects.

Seventy-two subjects with subretinal or intraretinal fluid on optical coherence tomography (OCT) at the time of first injection were included in this analysis. The remaining patients were excluded as follows: (1) another treatment for neovascular AMD was given within 3 months prior to the first injection with bevacizumab or ranibizumab (n = 17); (2) there was no history of treatment with bevacizumab or ranibizumab (n = 24); (3) injections were given at another institution during the first 12 months of treatment (n = 31); (4) no intraretinal or subretinal fluid was seen on OCT at the time of first injection (n = 13); or (5) a subject had incomplete data at the time of any treatment or follow-up visit, or less than 12 months follow-up (n = 53). Only the first eye treated with anti-VEGF intravitreal injection was included in the analysis. Best-corrected Snellen VA and central foveal thickness as measured by OCT imaging, performed on either the Stratus or Cirrus OCT (Carl Zeiss Meditec, Inc, Dublin, California, USA), were recorded for each visit in which treatment was received, and each post-treatment follow-up visit for a period of 1 year, as well as at 6-month and 12-month visits. Central foveal thickness measurements were recorded as the central value on the retinal thickness map, calculated by standard OCT retinal mapping software. Diagnosis of neovascular AMD was made by a retina specialist on the basis of a dilated fundus examination, OCT findings, color fundus photography, and fluorescein angiography. Initial and subsequent anti-VEGF treatments were given at the discretion of the treating physician, and were based on VA, a dilated fundus examination, and OCT results during follow-up visits. Demographic (age, sex, and education) and behavioral (BMI and smoking history) data were also evaluated. Genotypes were derived from genotyping and sequencing platforms as previously described. The following AMD-related genes were assessed as risk factors in these analyses: CFH , CFB , C3 , CFI , VEGFA , TGFBR1 , LIPC , ABCA1 , CETP , APOC1/APOE , TNFRSF10A , SLC16A8 , COL8A1 , COL10A1 , ARMS2/HTRA1 , RAD51B , DDR1 , ADAMTS9/ADAMTS9-AS2 , and HSPH1/B3GALTL .

Literature Review

A PubMed search using the terms “age related macular degeneration genetic response to treatment” was conducted at the time of the study. English language research articles assessing the association between genetic variables and response to intravitreal anti-VEGF treatment that were available before October 16, 2014 were included in the review. A Google Scholar search was also conducted using the same terms and inclusion criteria. Review or meta-analysis articles, as well as those that did not evaluate genetic risk factors, were not included. Details of these studies are included in Supplemental Table 1 (Supplemental Material available at AJO.com ).

Statistical Analysis

VA measurements at baseline, 6 months, and 12 months were assessed. OCT central foveal thickness measurements were analyzed at baseline and either 6 or 12 months follow-up. Snellen VA was transformed to logarithm of the minimal angle of resolution (logMAR) units by the following calculation: log (VA)/20. Means and standard deviations of log (VA)/20 and OCT central foveal thickness were calculated for each time point. Higher mean values of log (VA)/20 were indicative of worse VA. Mean change and standard error using Early Treatment Diabetic Retinopathy Study (ETDRS) acuity in letters were also calculated.

Longitudinal analyses were conducted to examine the association between demographic, behavioral, and genetic factors and change in VA and central foveal thickness measurements after anti-VEGF treatments. The associations between risk factors and the 2 outcomes over time were analyzed using mixed-effects linear regression models. Improvement in these outcomes was assessed based on slope, or change over time. A negative slope indicates improvement over time, a positive slope indicates worsening, and a slope of zero is equal to no change. Statistical significance was assessed using standard P values and P values for heterogeneity ( P _het ). P values were obtained to assess whether the change for each level of a variable was significantly different from zero (ie, no change) for both VA and central foveal thickness. P _het assessed whether the change over time for distinct levels within one particular risk factor significantly differed from each other for a particular outcome.

Changes in VA and central foveal thickness over time were also evaluated according to risk score category. These analyses used an externally derived and previously calculated composite score to assess risk of progression to advanced AMD. The risk score was calculated using regression coefficients of all demographic, behavioral, genetic, and ocular factors. The hazard ratio for the ith subject is given from the Cox proportional hazards model by <SPAN role=presentation tabIndex=0 id=MathJax-Element-1-Frame class=MathJax style="POSITION: relative" data-mathml='λi=exp∑j=1Jβjxij,’>λi=exp(∑Jj=1βjxij),λi=exp∑j=1Jβjxij,
λ i = exp ∑ j = 1 J β j x i j ,
where β _j is the regression coefficient for the jth variable and x _ij is the value of the jth variable for the ith subject. The variables included in the score are as follows: 10 genetic loci, age, sex, education, BMI, and smoking status. The low-risk category was defined as less than the median score, and the high-risk category was defined as greater than or equal to the median risk score.

A risk score for complement factor H ( CFH ) was calculated using the number of risk alleles for selected single nucleotide polymorphisms (SNPs). The score was composed of rs1061170 and rs1410996, with subjects having 0–4 risk alleles for these selected CFH variants. Groups for each score were classified as low risk (0–2 risk alleles) or high risk (3–4 risk alleles).

We also calculated similar risk scores for complement factor I ( CFI) , as variants in CFI have been implicated as potentially related to outcome after lampalizumab treatment in a small study (Regillo CD. Lampalizumab [anti-factor D] in patients with geographic atrophy: the MAHALO phase 2 results. Paper presented at the 2013 Annual Meeting of the American Academy of Ophthalmology, November 16-19, 2013, New Orleans, LA.). The CFI risk score comprised the number of risk alleles for CFI rare variants (0–2). The common CFI variant, rs10033900, was also evaluated based on number of risk alleles (0–2). CFI risk was classified as low or high risk by the number of risk alleles present (0 and 1–2, respectively).

As this study is considered exploratory, multiple comparisons and interactions between risk factors were not assessed. All analyses were conducted using SAS 9.3 (SAS Institute, Inc, Cary, North Carolina, USA). The alpha level for statistical significance was set a priori at P < .05.