In the present study, age, baseline VA score, duration of neovascular AMD, and ARMS2/HTRA1 genotypehave been identified as the predictors of the visual response to bevacizumab treatment at 6 months. Baseline VA score and ARMS2/HTRA1 genotype were both associated with the two measurements of visual response, the VA score change (letters), and the proportion of those gaining ≥15 letters. Cigarette smoking was found to be associated with an OCT feature: central retinal thickness (CRT) change.
The MARINA and ANCHOR studies have both shown that baseline VA was the most important predictor of VA outcome with ranibizumab treatment [19
]. Similar to those studies, we also identified baseline VA as the most influential predictor of VA outcome with bevacizumab treatment. Consistent with the CATT study, we found that the worse the baseline VA of the eyes, the more significant improvement in VA [21
]. One explanation may be that in the patients with the best baseline VA vision cannot be completely restored to what it had been before the presence of CNV, so their change in VA may be less than that of the patients with a worse baseline VA. This idea is supported by the inverse correlation, found in the current study, between the baseline VA score and the proportion of those with ≥15 letters gained. The association between baseline age and VA score change found in our study is also consistent with previous findings [19
]. According to the multivariate analysis, the change in baseline VA decreased as age increased. We found a decrease in the change from baseline VA of ~3 letters for every 10 years of age. However, the negative results of age with the proportion of those gaining ≥15 letters may have resulted from information loss after transforming a continuous variable into a categorical variable. Moreover, 6 months is not a long duration; thus, it is possible that the outcome measurement of ≥15 letters gained will have a higher statistical efficiency in longer-term data analysis.
The duration of neovascular AMD is a new predictor that was identified in the present study. However, two previous studies failed to confirm any association between the duration of neovascular AMD and the VA outcomes [19
]. Different inclusion criteria may be one reason for the difference. Compared with the previous studies, the eligible baseline VA criteria of the NATTB study were broader (between 5 letters and 73 letters), so more patients were considered to have experienced a longer duration of AMD in this study. In other words, the different composition of participants may have led to the different results. This finding implies that patients should receive appropriate treatment once diagnosed with a CNV.
The relationship between ARMS2/HTRA1
and VA score change from baseline to the 6th month is consistent with the previous analysis of visual outcome measures at 3 months in the NATTB study [24
]. Furthermore, the effect of the genes on the visual outcomes was independent of other predictors. We revealed that eyes with the TT genotype in ARMS2 rs10490924
or with the AA genotypes in HTRA1 rs11200638
had a smaller improvement in VA. Several previous studies did not find the association between the response to anti-VEGF therapy (ranibizumab or bevacizumab) and the genotype in ARMS2 rs10490924
], however; 2 other studies have reported that HTRA1
may influence the response to treatment of neovascular AMD with ranibizumab [31
]. As one of the most important susceptibility genes for AMD, the mechanism of how the ARMS2/HTRA1
genes influence the occurrence and development of AMD has been widely studied, and their association with the response to anti-VEGF treatment is still controversial. More research is needed on this mechanism in order to elucidate the impact of the ARMS2/HTRA1
genes. The association of the CFH rs800292
genotype with the ≥15 letters gain from baseline at 6 months was significant. However, it was found to be negative in the multivariate model as this genotype did not appear to be associated with a change in VA score from baseline to the 6th month; this finding was inconsistent with the results of the 3-month data analysis [24
]. More short- and long-term observations on the association between CFH rs800290
and the response to bevacizumab treatment are needed to resolve this inconsistency.
Cigarette smoking, which has been identified as the strongest environmental risk factor for the development of AMD, also showed an association with the change in CRT from baseline to the 6th month. In vitro
and in vivo
, nicotine has been found to upregulate the expression of VEGF and was reported to be responsible for the increase in the VEDF/PEDF ratio in RPE cells [33
], and oxidative injury induced by hydroquinone, a major prooxidant in cigarette smoke might lead to increased expression of VEGF protein and decreased expression of PEDF protein [35
]. Furthermore, dioxin which is present primarily in the gaseous phase of cigarette smoking promotes VEGF production in the retina of mice and human retinal pigment epithelium (RPE) cells and exacerbates the development of laser-induced CNV [36
]. Cigarette smoking plays a role in the pathogenesis of neovascular AMD in more than one way, including causing oxidative damage [37
], as well as affecting choroidal blood flow [39
] and macular pigment optical density [41
]; therefore, smoking may have an antieffect of bevacizumab treatment on the pathological characteristics in the retina for neovascular AMD, and, as a distant factor for retina, smoking may cause the anatomical change at first and then lead to dysfunction of retina, which may explain why cigarette smoking showed no association with VA score change that was a measurement of visual function.
Limitations of this study include that people in the cohort were not treated equally with bevacizumab and its small sample size. Although it was seen that there was significant difference in letters gained between treatment group A and group B in the 6th month data analysis, the patients in group A were actually treated only one more time compared with the patients in group B. Moreover, treatment regimen variable was not retained in the multivariate model, so the difference between the two groups in the 6th month data might result from other important predictors such as baseline VA score. The results were even similar in the two groups after association analyses between demographic and eye characteristics and VA score change from baseline at 6 months were conducted in the two regimen groups, respectively. So, the data of the two groups were combined to be analyzed in consideration of power issue. And further study might compare factors influencing the visual outcomes between the two groups if increasing the cohort size.
In summary, based on the analysis of the 6-month data from the NATTB study, we found that age, baseline VA score, duration of neovascular AMD, and ARMS2/HTRA1 genotype were all independent predictors of VA score change. Of these, baseline VA score was the most important predictor of visual response at 6 months following bevacizumab treatment of neovascular AMD. Cigarette smoking was found to decrease the improvement of CRT. Analysis of subsequent follow-up data may reveal a long-term effect of anti-VEGF treatment at different levels of these predictors.