Ranibizumab non-response in pachychoroid neovasculopathy: Effects of switching to aflibercept

Non-response to intravitreal ranibizumab represents a frequent problem in pachychoroid neovasculopathy (PNV). To investigate the effectivity of switching to aflibercept, the database of the Ludwig Maximilians University, Munich, was screened for patients fulfilling the following inclusion criteria: (i) diagnosis of PNV; (ii) inadequate response to ≥ 3 ranibizumab injections, in spite of monthly dosing, defined as persistence of subretinal-fluid four weeks after the last ranibizumab injection; (iii) resulting switch to aflibercept administered as three monthly injections. Primary outcome measure was percentage of eyes with a dry macula four weeks after the third aflibercept injection. Secondary outcome measures included changes in maximum subretinal fluid (SRF), central subfield thickness (CST) and subfoveal choroidal thickness (SFCT). In total, 14 eyes of 14 patients were included. Mean age was 64.1 ± 7.5 (range: 51–78) years. Switching to aflibercept was performed after mean 8.4 ± 4.1 (3–15) ranibizumab injections. While no eye (0%) achieved a dry macula status during ranibizumab treatment, switching to aflibercept achieved a dry macula status in eight eyes (57.1%) after three injections. While both ranibizumab and aflibercept showed an effect on CST (p = 0.027, p = 0.003), only aflibercept showed a significant effect on SRF (p = 0.0009) and SFCT (p = 0.044). In cases of PNV not responding to intravitreal ranibizumab, switching treatment to aflibercept induces a favorable short-term response resolving persistent fluid and achieving a dry macula. Further studies with longer follow-up are warranted.


Multimodal imaging.
Multimodal imaging (all on Spectralis HRA + OCT, Heidelberg Engineering, Heidelberg, Germany) was performed after pupil dilation with topical tropicamide 1% and phenylephrine 2.5%. It included enhanced depth (EDI) spectral domain optical coherence tomography (SD-OCT) and near-infrared (NIR) confocal laser scanning ophthalmoscopy (CSLO) in every eye at each visit. EDI SD-OCT was acquired using the volume mode in the high-speed setting with 49 B-scans (512 ×496) covering an area of 20° x 20° centered on the macula with the EDI mode "on". OCT angiography was performed in every eye at baseline. During follow-up, the presence of lessions suggestive of PCV/AT1 was ruled out using OCT, OCT angiography and fluorescein/indocyanine green angiography (FA/ICGA) in all eyes. Blue-autofluorescence (BAF) CSLO and additional OCT angiography scans were performed at the investigator's discretion.

Measurement of subretinal fluid (SRF), subfoveal choroidal thickness (SFCT) and central subfield thickness (CST).
Measurements of SRF and SFCT were obtained using the Heidelberg Eye Explorer (Heidelberg Engineering, Heidelberg, Germany) on enhanced depth imaging OCT images in the 1:1 µm setting. SRF thickness was measured at its maximum height from the outer portion of the photoreceptors to the retinal pigment epithelium. SFCT was measured directly underneath the fovea from the outer portion of the retinal pigment epithelium to the sclerochoroidal interface. Automated CST measurements were directly extracted from the software. Where needed, segmentation was manually adjusted.
Anti-VEGF treatment. All eyes were treated with monthly anti-VEGF injections. Treatment was started on-label with three monthly injections of ranibizumab (Novartis Pharma AG, Basel, Switzerland). A first evaluation of the treatment response was performed four weeks after the third injection. In the case of persistent macular fluid, treatment could be continued with monthly ranibizumab or switched to monthly aflibercept. Statistical analysis. All data were gathered and analyzed in Microsoft Excel spreadsheets (Version 16.23 for Mac; Microsoft, Redmond, WA, USA). Statistical analysis was performed in SPSS Statistics 25 (IBM Germany GmbH, Ehningen, Germany). The level to indicate statistical significance was defined as p < 0.05. The Shapiro-Wilk and Kolmogorov-Smirnov tests were employed to test for normal distribution. Statistical analyses of intra-group differences were performed using the dependent two-tailed Student t-test and the Wilcoxon signed rank test. A repeated measures ANOVA test was used to compensate for multiple testing, if applicable. Pearson's correlation coefficient was used to test associations of dependent and independent variables.

Macular morphology.
At baseline, all eyes (100%) showed subretinal fluid and had a flat, irregular PED forming a double-layer sign. The presence of a type 1 CNV within the flat PED was confirmed on OCT angiography in all eyes (100%). At baseline, no eye showed intraretinal fluid, a serous pigment epithelium detachment or a type 2 CNV configuration. None of the partner eyes showed soft drusen or reticular pseudodrusen/subretinal drusenoid deposits.
Primary outcome: percentage of eyes with a dry macula. None of the 14 eyes (0%) achieved a dry macula status during the ranibizumab treatment period, as all showed persistent subretinal fluid at every visit. After the third injection of aflibercept, eight eyes (57.1%) achieved a dry macula status without SRF, IRF or serous PED ( Fig. 1 and 2). In these eyes, mean 1.7 ± 0.9 (1-3) aflibercept injections were performed prior to complete fluid resolution. One eye (7.1%) achieved a dry macula status after the first injection of aflibercept, however showed recurrent SRF after the second and third injection.

Adverse events.
No serious adverse events (no endophthalmitis, no retinal detachment, and no macular hemorrhage involving the fovea and requiring pneumatic displacement) were observed during the study period.

Discussion
The present study was conducted to test the short-term response of intravitreal aflibercept in the treatment of PNV not adequately responding to intravitreal ranibizumab, defined as persistent macular fluid after ≥ 3 ranibizumab injections. According to recent data, such non-response in PNV might affect up to 50% of patients 7 .
In spite of a previous treatment with mean 8.4 ranibizumab injections, no eye included in this study showed a dry macula status before switching. In accordance, both subretinal fluid as well as SFCT did not show a significant decrease during ranibizumab treatment. Strikingly, switching treatment to aflibercept administered in three monthly loading doses however led to a significant decrease in sub-retinal fluid and SFCT, resulting in a dry macula status in 57% of eyes, in spite of the prior complete ranibizumab non-response.
In this study, and generally in PNV, the question of disease activity and resulting treatment decisions are mainly based on the presence of subretinal fluid. In contrast to other neovascular diseases associated with CNV, e.g. AMD or myopia, subretinal fluid in pachychoroid disease can however pose a diagnostic conundrum as its presence can equally result from neovascular activity, i.e. CNV exudation, or the pachychoroid disease itself, defined by choroidal hyperpermeability and leakage 2 . At the moment, no imaging modality has been shown to safely differentiate between subretinal fluid originating from CNV or the pachychoroid.
Thus, two reasons could explain why aflibercept showed better fluid resolution than ranibizumab in our study. On the one hand, more probable according to our data, it might be possible that CNV growth and leakage was better addressed by aflibercept. This could be due to its higher binding affinity to VEGF-A, or due to its broader    Table 2. Treatment outcomes of switching from ranibizumab to aflibercept. www.nature.com/scientificreports www.nature.com/scientificreports/ mechanism of action, which also includes Placental Growth Factor (PlGF) 17,18 . Of note, recent data suggest that VEGF might not play such a prominent role in PNV as in neovascular AMD, and thus, other growth factors, e.g. PlGF, might be more involved 19 . Thus, superior CNV inactivation by aflibercept might secondarily induce choroidal thinning, as leakage from the CNV lesion into the choroid, maybe even choroidal stroma, might be reduced. Indeed, Invernizzi et al. have shown that choroidal thickness and choroidal vascularity index are highly correlated with activity of CNV in neovascular AMD 20 . In turn, successful CNV inactivation by aflibercept might be more effective in a secondary choroidal thinning, and thus reduce the disease burden more effectively.
On the other hand, choroidal thinning might hypothetically also result from superior anti-VEGF action of aflibercept on the choroid. While both ranibizumab and aflibercept have proven to be effective in the treatment of PNV 7,8,21 , Jung et al. 7 have indeed recently suggested that aflibercept might be more effective on PNV due its stronger effects on the choroid. In their study of 54 PNV eyes, they found that a dry macula status was achieved in 82.6% of aflibercept, and 51.6% of ranibizumab treated eyes after three loading doses, an effect which seemed to be largely explained by choroidal features. Moreover, Kim et al. recently found that aflibercept induced more choroidal thinning in eyes with neovascular AMD than ranibizumab; strikingly, this effect was strongest in eyes  www.nature.com/scientificreports www.nature.com/scientificreports/ with an underlying pachychoroid etiology, i.e. polypoidal choroidal vasculopathy/pachychoroid aneurysmal type 1 CNV 10 . As stated above, as a major caveat, choroidal thinning might however also result from CNV or branching vascular network/polyp inactivation. Clinicians should bear in mind that anti-VEGF has not been shown to act successfully on acute CSC and its underlying pachychoroid 22 . Therefore, additional studies are needed to investigate whether response to different anti-VEGF substances in PNV is related to choroidal responses.
In our study, ranibizumab only showed an effect on CST, but not on SRF. As CST incorporates the distance from the ILM to Bruch's membrane (including retinal thickness, intra-/subretinal fluid and pigment epithelium detachments), this indicates that ranibizumab largely modulated PED height including the CNV, and to a lesser extend SRF, which thus might mostly result from the pachychoroid. Unfortunately, due to the retrospective nature of our study, we lack longitudinal repeated indocyanine green angiograms to prove this hypothesis by analyzing differential effects of both substances on the severity of choroidal leakage. Moreover, future longitudinal analyses of CNV morphology on OCT angiography before and after switching treatment might give better evidence if vascular remodeling (e.g. CNV size) is better achieved by one substance or the other. As an indirect measure, we were however able to show that PNV non-response to ranibizumab, defined as persisting subretinal fluid, was associated with a stable choroidal thickness -which, in other words, did not show any influence of ranibizumab. In contrast, choroidal thickness however significantly decreased after switching to aflibercept -a finding that was in turn associated with a resolution of sub-retinal fluid in more than half of the eyes included. At this moment, we do not know whether this is due to the superiority of aflibercept in CNV inactivation, or its effects on the choroid.
Certain limitations to our study can be found. One lies within the lack of ICG at baseline, which makes the standard assessment of polyps/aneurysms difficult. As a substitute, we however performed an in-depth analysis of typical OCT and OCT angiography to rule out the diagnosis of PCV/AT1; both OCT and OCT-A alone have been shown to produce a negative predictive value of more than 0.95 13,14 . Thus, the presence of undetected polypoidal/aneurysmal lesions at baseline in our cohort cannot be completely ruled out, but seems highly unlikely; as ICG as gold standard was however only performed during follow-up, a very small probability of polyps/aneurysms being present at baseline regressing during anti-VEGF therapy remains. Moreover, our study has a limited sample size, which reduces our statistical power, and the retrospective nature of this report does not allow for a comparison with a control group continued on ranibizumab. As pointed out by Ferris et al. in a report on the effects of anti-VEGF treatment switching in neovascular AMD and diabetic macular edema 23 , a significant amount of patients primarily presenting with suboptimal treatment responses in the beginning can show good late anatomical and functional outcome after a prolonged period of treatment, which might falsely increase the number of patients supposedly "finally reacting" to a treatment switch. Regarding our data, we however believe that this aspect might be of lesser importance as the cohort reported in this study received a rather high amount of mean 8 ranibizumab injections prior to switching, and anatomical outcomes were already evaluated after a short follow-up of only three injections of aflibercept. On the other hand, this short follow-up represents a limitation in assessing final visual acuity after switching, as a prolonged therapy with aflibercept might allow for a better reorganization of retinal layers once a dry macula is achieved. Moreover, even if we report anatomical success, no improvement in mean visual acuity could be found -which might be due to the cumulative damage caused by CSC previous to its neovascular conversion to PNV. In this context, sub-retinal fluid might either be secondary in its influence on visual acuity, or other end points beyond visual acuity, e.g. microperimetry might be more suited to estimate the translation of improved anatomy into improved function.
In conclusion, this study demonstrates that switching to aflibercept induces a favorable short-term response in cases of PNV not responding to previous intravitreal ranibizumab. Further prospective randomized clinical trials with a longer follow-up are warranted.