Introduction

Age-related macular degeneration (AMD) is the most common disease causing irreversible visual loss in patients of the industrialized countries over the age of 50 years [1]. The neovascular (wet or exudative) form is the most devastating configuration of AMD. Its main characteristic is the neovascularization process into sub-retinal and/or sub-retinal pigment epithelium (sub-RPE) spaces with consequent exudation and bleeding, followed by the formation of a scar in the macular region with loss of the central vision [2]. The emblem of the neovascular form is the choroidal neovascularization (CNV), and the vascular endothelial growth factor (VEGF) plays a key role in its pathogenesis.

Two landmark clinical trials, using intravitreal injections of an anti-VEGF-A antibody fragment ranibizumab in eyes with new-onset neovascular AMD, showed that over 90 % of treated eyes lost fewer than 15 Early Treatment Diabetic Retinopathy Study (ETDRS) letters after 2 years of monthly treatment, and the visual acuity could be improved in 33.3–40.0 % of treated eyes [3, 4]. Moreover, a recent randomized, multicenter, 24-month clinical trial revealed that bevacizumab, an anti-VEGF-A monoclonal antibody, was not inferior if compared to ranibizumab for the treatment of wet AMD [5].

However, despite the good results obtained, the CATT study showed that optical coherence tomography (OCT)-detectable macular fluid was present in 53.2 and 51.5 % of patients with monthly ranibizumab therapy after 1 and 2 years of therapy, respectively [5, 6]. Less frequent injections, pro re nata (PRN) protocol or bevacizumab use resulted in even greater percentages of patients with persistent retinal fluid [5]. This persistent macular fluid may be responsible for the ongoing visual impairment [5].

Aflibercept (or VEGF Trap-Eye) is a new anti-VEGF drug approved by FDA on November 18, 2011 [7]. It represents a recombinant decoy fusion protein with the two extracellular binding domains of VEGF receptors 1 (domain 2) and 2 (domain 3), neutralizing the biological activity of the pro-angiogenic factors VEGF-A, VEGF-B and placental growth factor (PlGF), which are present in the retina [8]. The presence of the extracellular regions of both VEGF receptors enables aflibercept to have a wider binding capacity than either bevacizumab or ranibizumab. Moreover, it has been hypothesized that this pharmacological characteristic could allow a more persistent, long-lasting VEGF blockade [8–11].

The VEGF Trap-Eye: Investigation of efficacy and safety in wet AMD (VIEW) 1 and VIEW 2 studies demonstrated that aflibercept, given every 2 months after a loading phase of 3 monthly intravitreal injections, was not inferior to monthly ranibizumab in maintaining visual function in patients with treatment-naïve neovascular AMD over 1-year period [7]. These results were maintained over the 96 weeks of the study [12].

However, information on long-term functional outcomes of anti-VEGF therapy in neovascular AMD is few and scattered because they are limited by the decrease in patient population in follow-up. A 7-year follow-up study on ranibizumab showed a mean vision loss ≥8 ETDRS letters [13]. Moreover, 34 % of patients had a loss of at least 3 lines in vision in the final visit [13]. The authors concluded that the vision loss in long-term follow-up is probably multifactorial [13]. Indeed, this vision loss could be due to pathological processes and pharmacological determinants such as the natural progression of the underlying non-neovascular AMD, too few injections of anti-VEGF drugs despite a still active disease, and, above all, the loss of treatment effectiveness over time [14]. Several studies have shown that a reduced anatomical response can be found for both ranibizumab and bevacizumab in neovascular AMD over time [15, 16]. Currently, such long follow-up has not been reported for aflibercept.

Different terms have been used to describe this clinical finding, ranging from tolerance or tachyphylaxis to resistance or not (insufficient or low) responsiveness [14, 17]. In general, in case of chronic administration of medications the decrease in pharmacological effect is a major concern that can limit the long-term efficacy. This phenomenon has been termed tachyphylaxis, referring to a progressive decrease in therapeutic response after repetitive administrations of a pharmacologically active substance [18]. On the other hand, innate resistance may be defined as cases that do not respond well functionally and/or morphologically since the first administration of the drug [19].

To overcome both innate resistance and tachyphylaxis, several treatment strategies have been proposed, such as switching to other anti-VEGF therapies [20] or increasing the concentration of the drug [21, 22] or the frequency of the injections [9]. Indeed, neovascular AMD patients refractory to standard ranibizumab doses (0.5 mg/0.05 ml), who were treated with higher dosage (ranibizumab 2.0 mg/0.05 ml), showed both anatomical and functional response [21, 22].

Gasperini and colleagues reported a significant decrease in sub-retinal fluid in the majority of tachyphylactic patients after switching treatments from either ranibizumab or bevacizumab to bevacizumab or ranibizumab, respectively [20]. The study showed that 81 % of the patients responded favorably after changing intravitreal injection medication, even if functional success was limited [20].

In this review, we specifically examine published articles reporting the influence of switching therapy from bevacizumab and/or ranibizumab to aflibercept in patients affected by resistant neovascular AMD. We hypothesized possible mechanisms responsible for the onset of tachyphylaxis or the presence of innate resistance to current available therapies. Moreover, treatment options to manage neovascular AMD patients refractory to “old” anti-VEGF drugs are proposed.

Methods

Search strategy and selection criteria

Data were identified by searches in Medline, Cochrane database, Current Contents, PubMed, and cross-referencing from identified articles; some articles were identified through searches of the extensive files of the authors. Search terms for the online research were as follows: “intravitreal aflibercept and age-related macular degeneration,” “VEGF Trap-Eye,” “VEGF Trap-Eye and age-related macular degeneration,” “aflibercept,” “aflibercept and macular diseases,” “Eylea,” “Vascular Endothelial Growth Factor Trap-Eye,” “insufficient responders to anti-VEGF,” “not responders to anti-VEGF,” “recurrent wet AMD,” “resistant wet AMD,” English language papers were reviewed.

All cases of “naïve AMD patients” were excluded. We reviewed the available scientific literature regarding the treatment of exudative AMD patients refractory to bevacizumab and/or ranibizumab and switched to aflibercept monotherapy. We included in this review all the cases in which the diagnosis of refractory or resistant exudative AMD was properly made, and the results of at least one aflibercept injection were described. Published literature has been reviewed until February 17, 2015.

Cases of polypoidal choroidal vasculopathy (PCV) were excluded from the analysis [23–27]. PCV is characterized by polyp-like lesions at the end of choroidal vessels. PCV is a common type of exudative AMD in the Asian population [28, 29]. Although typical AMD and PCV have been classified in the same category of exudative AMD [30], some of the PCV cases have the unique characteristic of choroidal vascular hyperpermeability [31, 32]. Moreover, the efficacy of intravitreal ranibizumab to treat PCV has been described to be lower than that for typical AMD [33–35]. In particular, PCV cases with choroidal hyperpermeability seem to be not related to VEGF, and thus, they may not respond favorably to anti-VEGF monotherapy [36].

Ho et al. reported the results of 96 eyes, one of which was a PCV, but they did not specify the outcome of this single eye, and consequently, it was not possible to separate this case from the overall analysis [37]. They also included 5 eyes (5.2 %) previously treated with verteporfin photodynamic therapy (V-PDT) and 1 eye with pegaptanib sodium injections [37]. Also, Kumar et al. [38] included in their analysis 5 eyes (14.71 %) previously treated with V-PDT, while Grewal et al. [39] included only 1 patient previously treated with one PDT. Finally, Yonekawa et al. [40] analyzed 6 eyes (5.88 %) previously treated with PDT, 1 eye (0.98 %) with thermal laser, and 2 eyes (1.96 %) with pegaptanib sodium. However, given the low percentages and the inability to extrapolate the individual results from the overall analysis, we decided to include these clinical studies in the analysis of the current review. Moreover, Kawashima et al. [41] reported the outcomes of both resistant neovascular AMD and PCV patients after switching the therapy from ranibizumab to aflibercept; however, they divided the analysis into two groups and reported outcomes from each group, so it was possible to extrapolate the outcomes from only neovascular AMD patients, which have been added in this review.

Also, the work of Fujii et al. [42] was not considered for the analysis because all the 3 cases of refractory nAMD reported in the article (one was a PCV) were complicated by retinal epithelial tear. Hanh reported a case in a vitrectomized eye, and it was excluded for this reason [43].

In the work of Rusu et al. [44], the authors focused their attention only on the modifications of intraocular pressure after switching the therapy, so it was not possible to include this article in our review because of the lack of clinical outcomes.

Broadhead et al. [45] analyzed the response of pigment epithelial detachment (PED) to intravitreal aflibercept among patients with treatment-resistant neovascular AMD in the same sample of the work of Chang et al., and it was excluded from our analysis. Finally, the case report of de Oliveira et al. [46] was excluded because they used ziv-aflibercept (Zaltrap; Sanofi-Aventis, Paris, France), which is different from aflibercept, because it is hyperosmolar (1000 mOsm/L) relative to the vitreous, while aflibercept is iso-osmolar.

Most of the reports were retrospective studies [37, 38, 40, 47–59], whereas only five were prospective trials [39, 41, 60–62].

Results

Using strict a priori criteria for this review, we identified a total of 21 papers. We selected two different groups: the first group included prospective works [39, 41, 60–62] and the second group included retrospective works [37, 38, 40, 47–59] . The results of our review are summarized in Tables 1, 2, 3, 4, 5, 6, 7, and 8.

Table 1 Prospective clinical trials: demographics
Full size table
Table 2 Retrospective clinical trials: demographics
Full size table
Table 3 Prospective clinical trials: previous anti-VEGF drugs and aflibercept treatment
Full size table
Table 4 Retrospective clinical trials: previous anti-VEGF drugs and aflibercept treatment
Full size table
Table 5 Prospective clinical trials: anatomical outcomes after aflibercept injections
Full size table
Table 6 Retrospective clinical trials: anatomical outcomes after aflibercept injections
Full size table
Table 7 Prospective clinical trials: functional outcomes and adverse events after aflibercept injections
Full size table
Table 8 Retrospective clinical trials: functional outcomes and adverse events after aflibercept injections
Full size table

Prospective reports

Demographics (Table 1)

A review of the literature revealed 5 prospective reports for a total of 157 eyes affected by exudative AMD previously treated with anti-VEGF intravitreal injections and switched to aflibercept intravitreal treatment (mean age 78.13 years; 65 males, 66 females and 26 gender unknown) [39, 41, 60–62]. Laterality was reported only in one paper (23 right eyes and 26 left eyes) [61], while in the other works it was not reported [39, 41, 60, 62]. Only Chang et al. [61] reported the lens status (32 phakic eyes and 17 pseudophakic eyes). Follow-up duration ranged from a minimum of 6 months [41, 60–62] to a maximum of 12 months [39] after the first aflibercept injection. Inclusion and exclusion criteria, as well as the definition of refractory or recurrent exudative AMD, differed from each paper as reported in Table 1.

Previous anti-VEGF drugs and aflibercept treatment (Table 3)

Chang et al. [61] and Kawashima et al. [41] enrolled only patients previously treated with ranibizumab 0.5 mg/ml monotherapy, and Wycoff et al. [60] included patients previously treated with ranibizumab 2.0 mg from the SAVE Study, whereas Singh et al. [62] and Grewal et al. [39] enrolled patients previously treated with ranibizumab 0.5 mg/ml and/or bevacizumab 1.25 mg/ml.

The mean number of previous anti-VEGF injections was 29.55 for patients, even if great differences about the number of injections were reported not only among different works, but also within the same study (i.e., Grewal reported a range from 6 to 74 [39]). Interestingly, 22 eyes have been previously underwent to at least one switch from ranibizumab to bevacizumab or vice versa, and 46 eyes were previously switched from ranibizumab 0.5 mg/ml to ranibizumab 2.0 mg/ml [60] while all the other patients received only monotherapy with ranibizumab 0.5 mg/ml (77 eyes) or bevacizumab (12 eyes). The mean number of anti-VEGF drug injections in the last 6 months prior of the switching to aflibercept has been reported only in the work of Chang et al. (mean of 5.0 ± 0.7 injections) [61]. The mean time interval between the last anti-VEGF injection and the first aflibercept injection was reported only by Wycoff et al. [60] and Chang et al. [61] (mean of 33 days in the first case and at least 30 days in the second one).

The mean number of aflibercept injections was reported in all papers except for Kawashima et al. (mean number 5.96, range 5.0–10.2) [39, 60–62]. Aflibercept injection protocols provided in all cases: a loading phase of 3 monthly injections [39, 41, 60–62], followed by bimonthly injections [41, 62], one mandatory dose at month 4 and PRN doses at months 3 and 5 [60], mandatory injection at weeks 16 and 24 [61], bimonthly injections in case of resolution of the edema or monthly in case of edema [39].

Anatomical outcomes (Table 5)

Mean baseline central macular thickness (CMT) was 335.34 µm, even if great differences were reported not only among different works (range of mean values 202.1 µm [41] to 448.4 µm [61]), but also within the same study (range 193–637 µm) [39]. After the first aflibercept injection, CMT has been reported only in one work [60] with a significant reduction (P < 0.05). Data were available for three works after a 3-month follow-up [39, 60, 61]: Mean reduction in CMT was statistically significant in all works (P < 0.05) [60, 61] except for Grewal and co-workers (P > 0.05) [39]. At the 6-month follow-up, data were reported for all works [39, 41, 60–62]. Mean reduction in CMT was statistically significant in all works (P < 0.05) [39, 41, 60–62]. Total mean change was −52.47 µm (range from −25.24 µm [39] to −89.5 µm [61]). Only Grewal et al. [39] reported data from the 12-month follow-up with a significant reduction in the CMT [292.71 ± 91.35 µm (P = 0.038), range 193–627 µm].

The maximum height of PED was analyzed only in two studies [39, 41]. Total mean baseline value was 227.87 µm (range from 167 µm [41] to 288.73 µm [39]); mean total change at the 6-month follow-up was −60.52 µm (range from −72.9 µm [41] to −43.13 µm [39]) with statistical significance reported in all works (P < 0.05). Data of other parameters and follow-up visits are reported in Table 5.

Functional outcomes (Table 7)

In three studies, best-corrected visual acuity (BCVA) was measured with ETDRS charts at 4 meters [60–62]: mean baseline value was 64.83 ETDRS letters. After 6 months, mean change was +4.14 ETDRS letters. Only Wycoff et al. [60] did not report any significant change at 6-month follow-up after the switching to aflibercept (P > 0.05), whereas Chang et al. [61] and Singh et al. [62] reported a significant change at the last follow-up visit (P < 0.001). Sub-analyses showed great differences in individual response (range, from −10 letters to >+15 letters [60–62]).

In the other two articles, BCVA has been reported as Logarithm of Minimum Angle of Resolution (LogMAR) [39, 41]. These authors showed not significant change in BCVA at the 6-month follow-up (P > 0.05) [39, 41].

Adverse events (Table 7)

Reported ocular adverse events were rare and were reported only in two works [60, 61]. Wycoff et al. [60] reported cataract progression in 3 patients (7 %); progression of geographic atrophy in 3 patients (7 %); atrial fibrillation in one patient (2 %); finally one patient (2 %) died related to complications of acute onset leukemia; while Chang et al. [61] extensive sub-macular hemorrhage in 1 patient (2 %); worsening of a previous sub-macular hemorrhage after first injection in one patient (2 %); progression of cataract in one patient (2 %); acute myocardial infarction in one patient (2 %); deep vein thrombosis in one patient (2 %); atrial fibrillation in 2 patients (4 %); syncope in 1 patient (2 %); spondylolisthesis in 1 patient (2 %), gastritis in 1 patient (2 %); pneumonia in 1 patient (2 %).

Retrospective reports

Demographics (Table 2)

A review of the literature revealed a total 16 retrospective reports for a total of 909 eyes affected by exudative AMD previously treated with anti-VEGF intravitreal injections and switched to aflibercept intravitreal treatment (mean age 78.66 years; 331 males and 530 females) [37, 38, 40, 47–59]. Laterality was reported only in six papers (136 right eyes and 126 left eyes) [40, 47, 48, 50, 52, 55], while in the other works it was not reported [37, 38, 49, 51, 53, 54, 56–59]. Only Arcinue et al. [55] reported the lens status (19 phakic eyes and 44 pseudophakic eyes). Follow-up duration ranged from a minimum of 1 month [59] to a maximum of 12 months [52, 53, 55] after the first aflibercept injection. Inclusion and exclusion criteria, as well as the definition of refractory or recurrent exudative AMD, differed from each papers, as reported in Table 2.

Previous anti-VEGF drugs and aflibercept treatment (Table 4)

All the authors of the included papers enrolled patients previously treated with ranibizumab 0.5 mg/ml and/or bevacizumab 1.25 mg/ml [37, 38, 48–58], except for Michalewsky et al. (only bevacizumab 1.25 mg/ml) [59], Yonekawa et al. [40] (ranibizumab 0.5 mg/ml, ranibizumab 0.75 mg/ml, ranibizumab 1.00 mg/ml, bevacizumab 1.25 mg/ml, bevacizumab 2.00 mg/ml) and Patel et al. [47] (ranibizumab 0.5 mg/ml, ranibizumab 1.00 mg/ml, ranibizumab 1.50 mg/ml, bevacizumab 1.25 mg/ml).

The mean number of previous anti-VEGF injections was reported in all papers [37, 38, 40, 47–54, 56, 57] except in 3 works [55, 58, 59]: a mean of 21.22 anti-VEGF injections were administered before the switch to aflibercept treatment. Also in this group, great differences about the number of injections were reported not only among different works, but also within the same study (i.e., Bakall et al. [48] reported a range of 6–74). 161 on 694 eyes (23.2 %) have been previously underwent to at least one switch from ranibizumab to bevacizumab or vice versa, while all the other patients received only monotherapy with ranibizumab 0.5 mg/ml (174 eyes) or bevacizumab (359 eyes). Data were not available from 4 works [38, 50, 55, 58]. The mean number of anti-VEGF drug injections in the last 6 months prior of the switching to aflibercept has been reported only in 3 works (mean of 5.33 injections) [48, 54, 57] and in the last 12 months only in 2 works (mean of 8.36 injections) [53, 57]. The switch from anti-VEGF injections and aflibercept injections was greatly variable among different works, but also within the same report there was not uniformity (i.e., range 20–422 days [49]).

The mean number of aflibercept injections was reported in all papers except for Arcinue et al. (mean number 4.27, range of mean values was 1.0–6.27). Aflibercept injections protocols was greatly different among the included papers, as reported in Table 4.

Anatomical outcomes (Table 6)

Mean baseline Central Macular Thickness (CMT) was 326.15 µm, even if great differences were reported not only among different works (range of mean values 228.6 µm [58] to 521 µm [59]), but also within the same study (i.e., Bakall et al. [48] reported a range of baseline CMT from 181 to 579 µm). After the first aflibercept injection, CMT has been reported only in five works [40, 49, 50, 52, 59] with a significant reduction in all cases (P < 0.05). Total mean reduction in these five works was 35.51 µm (range, from −23 µm [50, 52] to −75 µm [59]). Data were available for seven works after a 3-months follow-up [38, 48, 49, 51, 52, 54, 56]. Mean reduction in CMT was statistically significant in all works (P < 0.05) except for Hall and co-workers (P = 0.102) [52]. Total mean change was −79.27 µm (range, from −16 µm [52] to −121.9 µm [49]). At the 6-month follow-up, data were reported for ten works [38, 40, 48, 50, 52–55, 57, 58]. Mean reduction in CMT was statistically significant in all works (P < 0.05) except for Hall and colleagues (P = 0.099) [52]. Total mean change was −48.13 µm (range, from −16 µm [52] to −180 µm [54]). Finally, three works reported the 12-months follow-up data [52, 53, 55]. Mean reduction in CMT was statistically significant in all the reports (P < 0.05), with a mean reduction in −93.11 µm.

The maximum height of PED was analyzed only in five studies [38, 48, 51, 54, 58].Total mean baseline value was 280.64 µm (range, from 153 µm [48] to 425.1 µm [58]); mean total change at the 3-month follow-up was reported in four works [38, 48, 51, 54] and mean reduction was −45.43 µm (range, from −26 µm [48] to −101 µm [54]) with statistical significance reported in all works (P < 0.05). At the 6-month follow-up data were available for four works [38, 48, 54, 58] that showed a mean change of −80.17 µm from baseline and statistical power in all the works. Data of other parameters and follow-up visits are reported in Table 6.

Functional outcomes (Table 8)

In three studies, best-corrected visual acuity (BCVA) was measured with ETDRS charts at 4 meters [51, 54, 57]: All the authors did not report any significant changes at 3 [51, 54] and 6 months [54, 57] after the switching to aflibercept (P > 0.05).

In the most part of the enrolled articles, BCVA has been reported as logarithm of minimum angle of resolution (LogMAR) [38, 40, 48–50, 52, 53, 55–59]. Michalewsky and co-authors reported a significant increase in BCVA 1 month after the first aflibercept injection (P = 0.03) [59]; Hall and co-workers and Kumar and colleagues found a significant improvement in BCVA at the 6-month follow-up (P < 0.05) [38, 52], whereas in the other works there was not a significant change at the last follow-up visit [40, 48–50, 53, 55, 56, 58]. Also in these works, a great individual variability has been found, with gaining of >15 letters or loosing >15 letters. Finally, Ho et al. [37] and Patel et al. [47] measured BCVA in 20/x, without any significant change after the treatment with aflibercept (P > 0.05).

Adverse events (Table 8)

Reported ocular adverse events were rare and included one case of endophthalmitis (subsequently resolved by vitrectomy with restoration of visual acuity) [48], onset of focal areas of sub-macular hemorrhage in 4 eyes [37], and increasing of intraocular pressure (IOP) >25 mmHg in 2 patients [52].

No systemic adverse events were observed during the study period in all the enrolled papers [37, 38, 40, 47–59].

Discussion

Eyes with recalcitrant exudative AMD represent a substantial clinical burden. It is unknown why some eyes with neovascular AMD dry up anatomically with fewer intravitreal injections of anti-VEGF drugs, whereas up to half have OCT findings of disease activity even with continuous monthly therapy [5]. However, funduscopy presentation of neovascular membrane represents only the final result of a complex cascade of events that can be modulated at different steps; therefore, they may require different treatments from patient to patient [63]. The early identification of these cases is essential in order to maintain the costs of these therapies sustainable for the national health systems.

The authors of the enrolled papers have addressed the problem switching to a new anti-VEGF now available for the treatment of neovascular AMD. Indeed, VEGF-A is the only target of bevacizumab and ranibizumab, whereas aflibercept binds all the isoforms of VEGF-A and VEGF-B, as well as PlGF [8]. Furthermore, aflibercept has a different molecular structure than ranibizumab and bevacizumab [64] with a significantly higher binding affinity for VEGF than either bevacizumab or ranibizumab (about 100 times greater) [11, 65]. Recent mathematical simulations predicted that a single intravitreal injection of aflibercept 2.0 mg would last between 48 and 83 days (compared with 30 days of ranibizumab 0.5 mg) and thus should be efficacious in neutralizing VEGF longer and more effectively [9, 11].

The aims of this review are as follows: (i) to report anatomical and functional outcomes of switching from bevacizumab/ranibizumab to aflibercept previously described in the scientific literature, (ii) to hypothesize the possible pathophysiological mechanisms of the resistance and tachyphylaxis to anti-VEGF drugs, and (iii) to suggest possible clinical actions increasing the chances of success for such difficult cases.

We reported the outcomes of 21 papers for a total of 1066 eyes affected by exudative AMD resistant to previous anti-VEGF drug injections and switched to aflibercept injections. Enrolled reports were divided into two groups: 5 prospective reports and 16 retrospective reports. Outcomes were not easy to decipher because of the different criteria used in these studies, although the detailed analysis of these data provided really interesting insights.

Of note, about a quarter of the eyes included in this analysis [68 out of 157 (43.31 %) and 161 out of 694 (23.19 %) eyes in the prospective and retrospective groups, respectively] were subjected to a previous switch from bevacizumab to ranibizumab and/or vice versa, potentially reducing the effects of switching to aflibercept. However, this may be not the case because, as previously described, bevacizumab and ranibizumab have the same target (i.e., VEGF-A) [64], whereas aflibercept binds all the isoforms of VEGF-A and VEGF-B, as well as PlGF, with a different molecular structure and significantly higher binding affinity for VEGF than either bevacizumab or ranibizumab (about 100 times greater) [11, 64, 65]. These dissimilarities might produce a better success rate than the simple switch bevacizumab/ranibizumab or vice versa. Finally, a single intravitreal injection of aflibercept is longer lasting than ranibizumab and thus should be efficacious in neutralizing VEGF longer and more effectively [9, 11].

Also, the “washout period” between the last anti-VEGF drug injection and the first aflibercept injection represents an important factor that could influence the final results of switching to aflibercept (Tables 3, 4). However, inconclusive data can be extrapolated from the included papers because not all the authors reported these data and great variability was present. In our hypothesis, if aflibercept has been administered too early, its effect could be limited by the previous anti-VEGF drug injection, competing for the same target. As reported by mathematical model of Stewart and co-workers, the pharmacological effect of ranibizumab 0.5 mg lasts about 30 days; hence, the switch to aflibercept should be performed from 4 to 5 weeks after the last anti-VEGF drug treatment [9].

Most of the studies reported in this review used a loading phase of three monthly aflibercept (Tables 3, 4), while subsequent injections were administered with different protocols and with different intervals of time (range from 4 to 8 weeks) (Tables 3, 4). Chang et al. [61] and Ho et al. [37] demonstrated that when the injections were administered every 8 weeks, the effects of aflibercept may not last for this duration as suggested by the fluctuation pattern of CRT. However, this pattern does not seem to impact significantly on vision [61]. The CRT was reversible with reinjection of aflibercept, even if it is currently not known whether fluctuating sub-retinal or intraretinal fluid, or both, is associated with long-term visual impairment [7].

Kumar et al. [38] described that in patients with persistent fluid, despite the previous treatments with other anti-VEGF drugs, the continuation of the therapy with aflibercept beyond three injections could be needed to achieve an improvement in visual acuity. Indeed, these eyes responded differently if compared to treatment-näive eyes evaluated in the clinical trials where a significant anatomical and visual improvement occurred after three injections [7]. Only Hall et al. and Grewal et al. reported data at 12-month follow-up, showing a significant anatomical response (P = 0.012 and P = 0.002, respectively) (Tables 5, 6), but not a functional one (P = 0.836 and P = 0.5, respectively) (Tables 7, 8) [39, 52]. In all the other papers, follow-up lasts maximum 6 months after the first aflibercept injection (Tables 3, 4).

The great variability in number of previous anti-VEGF drug injections could affect the results. Indeed, we hypothesize that few anti-VEGF drug injections could mean a sub-optimal treatment, whereas an excessive number of injections without obtaining a complete fluid reabsorption could represent an inadequate treatment for a long period, leading to the maturation of the blood vessels within the neovascular lesions that becomes less dependent of the VEGF action. This great variability that we can observe in the enrolled papers is the same of the daily practice, and it opens an important question: Which is the “perfect number” to define the non-responsiveness to anti-VEGF drug injections and to take into consideration other therapeutic options? In all the included papers, the mean numbers of previously anti-VEGF drug injections were really big (29.55 injections/patients and 21.22 injections/patients in the prospective and retrospective groups, respectively). This aspect could have influenced the final outcomes. Monthly intravitreal ranibizumab injections showed maximum responsiveness after the third injection in the ANCHOR study [4], whereas in the EXCITE study the maximum responsiveness was achieved after the sixth injection [66], followed by CMT and BCVA stabilization. Therefore, the “perfect number” could be smaller than that reported in the papers included in this review and should be in the range between 3 and 6 injections of the same anti-VEGF drug: If there is not a complete fluid reabsorption or reduction in CMT of at least 100 µm from baseline, treatment is not effective and a different therapeutic approach should be taken into consideration.

The most plausible pathophysiological hypothesis for initial resistance could be related to the different levels of vitreous VEGF-A in each patient. This hypothesis is supported by studies that have found higher VEGF vitreous concentrations in patients with neovascular AMD compared with healthy controls [67] and higher VEGF vitreous concentrations being associated with a worse prognosis [68]. Indeed, anti-VEGF drugs work by blocking the VEGF-A protein. In the presence of higher vitreous VEGF-A concentrations, the standard dose of the commercially available anti-VEGF drug could not completely neutralize the vascular growth factor [63]. Increasing the injected drug dose appears as a valid alternative, but extends biological activity by only half-life time [69]. In the SAVE study, neovascular AMD patients refractory to commercially available anti-VEGF drugs (ranibizumab 0.5 mg/0.05 ml) were switched to higher intravitreal concentrations of the same drug (ranibizumab 2.0 mg/0.05 ml) [21]. Despite the good results obtained, this clinical trial is burdened by an important limitation: The enrolled patients were probably treated for months with an inadequate dose of ranibizumab. Therefore, the functional recovery was limited (+3.3 ETDRS letters at 3 months), and retinal fluid was still present in 70 % (45/64) of patients at the end of 2 years [21]. Interestingly, also the LAST trial has investigated the effects of ranibizumab 2.0 mg/0.05 ml in refractory exudative AMD, with significant improvement in BCVA (P < 0.001), even though the interpretation of these data is limited by the small number of enrolled patients (n = 9) [22]. However, it should be noticed that in naïve exudative AMD patients the HARBOR study failed to demonstrate any clinical advantage of 2.0 mg ranibizumab dose over the 0.5 mg ranibizumab dose [70].

No large studies have been conducted to specifically evaluate the increased frequency of treatment to more than the standard, monthly treatment regimens. Using pharmacokinetic modeling, Stewart et al. [71] evaluated whether dosing bevacizumab or ranibizumab every 2 weeks could be beneficial. Their results suggested that the increased trough binding activity achieved with the 2 weekly dosing could explain the improved results noted in some of their patients with persistent fluid. However, biweekly ranibizumab is not approved by Food and Drug Administration (FDA) and by European Medicines Agency (EMEA); moreover, this treatment regimen has significant cost implications and higher ocular risks. In addition, any regimen requiring appointments and potential treatment every 2 weeks may be difficult to follow by the elderly AMD patients, potentially affecting compliance and hence outcomes.

Another explanation for the innate resistance to anti-VEGF drugs could be represented by early up-regulation of pro-angiogenic factors other than VEGF-A, such as VEGF-B, PlGF, tumor necrosis factor α (TNFα), and/or down-regulation of anti-angiogenic factors, such as soluble vascular endothelial growth factor receptor-2 (VEGFR-2) or thrombospondins [63].

Currently, the mechanism of tachyphylaxis is unclear. It has been estimated that nearly 2 % of the patients may develop tachyphylaxis during the anti-VEGF therapy [15]. The lower pharmacological response over time was noticed to develop regardless the initial treatment with ranibizumab or bevacizumab. Some authors observed that in some patients a decreased clinical response occurred just after two anti-VEGF drug injections, whereas other patients did not show tachyphylaxis until they underwent ten or eleven injections [20]. In the article by Forooghian and colleagues, the median time to develop tachyphylaxis was 100 weeks, and the median number of intravitreal bevacizumab injections prior to establish tachyphylaxis was eight [72]. Additionally, Schaal and co-workers described that nearly three injections were required before the efficacy decreased to 50 % of the initial OCT response [16]. Although no conclusive data are currently available on the modulation of pro- and anti-angiogenic factors in the vitreous after anti-VEGF drug injections, the clinical drug resistance could be reasonably determined by up-regulation of pro-angiogenic factors other than VEGF-A, such as VEGF-B [73] and PlGF [74] and/or down-regulation of anti-angiogenic factors such as thrombospondins. Indeed, continuous VEGF blockade up-regulates the production of pro-angiogenic factors and overwhelms the effects of anti-VEGF agents [17, 75–77]. Finally, a recent work by Leveziel and colleagues seems to suggest that immunization against ranibizumab could be observed and may influence the clinical response [78]. Indeed, after intravitreal injection, a systematic immune response to VEGF inhibitors in patients’ serum as well as a local immune response from a compromised blood-ocular barrier may contribute to the formation of measurable, neutralizing antibodies to bevacizumab and ranibizumab. Such a response also may account for the occurrence of sterile uveitis after repeated injections with anti-VEGF agents [79, 80]. Therefore, the switch to aflibercept, a drug with a wider range of targets (i.e., VEGF-A, VEGF-B, and PlGF) associated with the fact that no immune response could be present for the new drug, theoretically may lead to sustained benefits in eyes refractory to bevacizumab, ranibizumab or both.

Finally, pharmacogenetic aspects may have an important role in the identification of non-responders and in the management of neovascular AMD patients. In particular, the certain identification of a clear association between genotypes or haplotypes of genes involved in the therapeutic response of neovascular AMD could be a prerequisite for any further study to establish effective anti-VEGF treatment regimens and dosing outsets [81, 82]. As an example, Kitchens and colleagues, based on the results of optical coherence tomography, described that patients who carried the LOC387715 A69S TT genotype were significantly more likely to be classified as a non-responder compared to those with the GG and GT genotypes (P = 0.00071) [83], whereas Korean patients harboring VEGF rs699947 AA genotype had an increased chance of good response to ranibizumab compared with other genotypes (P = 0.0071) [84]. Moreover, the same group described, for visual outcome measures, that VEGFA rs3025039 CC and CT genotypes were significantly associated with the lack of visual improvement after month 24 from the beginning of ranibizumab treatment compared to patients carrying TT genotype [85].

Thus, based on the growing pharmacogenetic data, it may be conceivable to personalize the anti-angiogenic therapy based on the genetic background, starting the administration of the proper anti-VEGF drug, rather than waiting for the treatment failure before choosing a different therapeutic approach. However, randomized multicenter pharmacogenetic studies on ranibizumab, bevacizumab, and aflibercept should be required before to answer to this important issue.

Conclusion

In conclusion, analysis of the papers reported in this review demonstrates that switching from bevacizumab/ranibizumab to aflibercept injections can improve outcomes successfully in refractory neovascular AMD patients. The mechanism for these effects is not yet completely understood.

However, based on these data and premises, standard criteria for the early identification of the subset of non-responder patients are urgently needed. It would be possible to switch to a different therapeutic approach, such as aflibercept injections, increasing the chances to preserve visual acuity. Indeed, the quality of scar tissue is another key aspect for the final prognosis: if a proper and earlier treatment is administered, a smaller scar in the retinal tissue will be found, resulting in a better functional recovery. In fact, most of the reports of this review highlighted a significant anatomical response after switching to aflibercept, but not a significant functional improvement. This could be the results of an inadequate treatment administered for months or years, which led to a macular scar formation, reducing margins of visual acuity improvement even in case of complete reabsorption of the intraretinal and/or sub-retinal fluid.

Larger studies and longer follow-up are needed to determine whether these anatomical gains and VA findings can be sustained.