To Treat or Not to Treat? Resolving the Question of Subretinal and Intraretinal Fluid in Age-Related Macular Degeneration: A Narrative Review

Abstract

Neovascular age-related macular degeneration (nAMD) is associated with considerable quality of life and economic burden. nAMD is characterized by pathological neovascularization, leading to the accumulation of retinal fluid. Intraretinal fluid (IRF) is a major contributor to vision loss and may predict response to treatment. In contrast, the role of subretinal fluid (SRF) is less clear. Nevertheless, complete resolution of retinal fluid accumulation is often stated to be a key goal of therapy for nAMD, even though some eyes may never achieve a fluid-free macula despite regular anti-vascular endothelial growth factor treatment. In this article, we review the current literature regarding the role of retinal fluid in nAMD disease outcomes and assess whether and when it may be beneficial to leave retinal fluid untreated. In this context, we highlight the importance of correctly identifying retinal fluid types in nAMD and avoiding confusion with other optical coherence tomography signs that may respond differently to therapy, such as subretinal pseudocysts. Current evidence shows that IRF is associated with poor outcomes and an increased risk of developing atrophy and fibrosis; resolution of this retinal fluid type should remain a treatment target. However, the literature around SRF indicates that low levels of this fluid type, potentially up to 150–200 µm in thickness, may be tolerated with minimal impact on vision, and that SRF could be protective against the development and progression of macular atrophy and fibrosis. Although mild SRF may be protective in nAMD, cause and effect between SRF and reduced or slowed atrophy has not yet been proven and requires further research. Treatment should be given for the most aggressive component; when both IRF and SRF are present, treatment should be given for IRF.

Plain Language Summary

Neovascular age-related macular degeneration (nAMD) is a progressive eye disease and a major cause of vision loss. In nAMD, new blood vessels grow into the light-sensitive layer at the back of the eye called the retina and can leak fluid. Fluid leaked into the retina is known as intraretinal fluid, and behind the retina is known as subretinal fluid. It is known that intraretinal fluid contributes to vision loss, but the impact of subretinal fluid is less clear. One aim of treatment for nAMD is to dry up retinal fluid, but in some eyes, this treatment is ineffective. This article summarizes previous research on identifying retinal fluid, the effect of retinal fluid on nAMD progression, and how the presence of retinal fluid impacts how patients respond to nAMD treatment. Findings from previous research showed that compared with eyes without intraretinal fluid, eyes with intraretinal fluid have worse vision, and this is more likely to decline further over time. Therefore, treatment should aim to dry up intraretinal fluid. Interestingly, the literature also suggested that a low level of subretinal fluid has little impact on vision and might even protect against nAMD worsening. The impact of mild subretinal fluid has not been proven yet and requires further study. Therefore, when both intraretinal fluid and subretinal fluid are present, treatment should be given for intraretinal fluid.

Key Summary Points

Intravitreal anti-vascular endothelial growth factor injections are the standard of care for neovascular age-related macular degeneration (nAMD). However, these injections are burdensome on patients and healthcare systems, so it is desirable to avoid unnecessary injections.

The relationship between retinal fluid and visual outcomes is complex, and complete resolution of all forms of retinal fluid may not always be necessary.

Intraretinal fluid (IRF) is a poor prognostic factor in nAMD and should be treated to resolution where possible.

Mild subretinal fluid (SRF) may be protective in nAMD and, as such, low levels of SRF (up to 150–200 µm in thickness) may be tolerated if IRF is absent or resolved, especially if the external limiting membrane is intact. However, this hypothesis requires further investigation and cannot yet be recommended in clinical guidelines. Thicker SRF requires treatment.

Treatment decisions regarding whether to tolerate mild residual SRF should be individualized and may be influenced by additional factors beyond retinal fluid, such as macular neovascularization type, fibrosis, atrophy, and patient preferences.

Introduction

Retinal diseases, such as age-related macular degeneration (AMD), are associated with a substantial health-related quality of life burden for patients, affecting activities of daily living, and emotional and psychosocial well-being [1]. They also represent a considerable global economic burden arising from healthcare costs and loss of productivity [1–3]. A meta-analysis highlighted that AMD is the third most common cause of visual impairment globally and is expected to further rise in prevalence [4, 5]. Approximately 8.4 million people had moderate or severe vision impairment due to AMD in 2015, with the number estimated to rise to around 288 million people by 2040 due to population aging [4, 5].

Neovascular AMD (nAMD), a progressive and degenerative form of AMD, is estimated to affect over 30–40 million individuals worldwide [6–8]. nAMD is characterized by pathological neovascularization in the subretinal spaces or subretinal pigment epithelium (sub-RPE) space, causing rapid and severe vision loss [6]. In the long term, the primary causes of vision loss in nAMD include atrophy and fibrosis, as well as leakage and hemorrhage. Long-term visual loss is due to permanent structural damage, which includes sub-retinal fibrosis and atrophy [9]. Leakage can be dried up with treatment and hemorrhage itself will resolve, but is a predictor for developing permanent structural damage [9, 10]. Macular neovascularization (MNV) is the ingrowth of new blood vessels and associated tissues into the outer retina, subretinal space, or sub-RPE space. MNV is a feature of nAMD that can lead to exudation of fluid, bleeding, and fibrosis, which each contribute to severe vision loss [11]. In addition, as well as potentially contributing to vision loss, the accumulation of fluid in and around the retina is also a predictor of treatment response [12]. In active nAMD, the retinal pigment epithelium (RPE) produces growth factors, including the pro-angiogenic vascular endothelial growth factor (VEGF); excess VEGF is a key driver of angiogenesis, leading to MNV [13]. These newly formed vessels may be dysfunctional and leaky as VEGF affects their permeability, leading to exudation and accumulation of fluid into the macula [13, 14].

A number of questions remain regarding the significance of retinal fluid, particularly when to treat it, when it should be monitored, and when, if ever, it might be beneficial to leave untreated. In this review, we explore the role of retinal fluid in nAMD and the potential usefulness of different types of retinal fluid biomarkers in guiding treatment regimens. The primary focus is on subretinal fluid (SRF) and intraretinal fluid (IRF), with the aim of addressing the remaining clinical questions.

Given the key role of VEGF in nAMD, intravitreal anti-VEGF treatment has become the standard of care approach for the treatment of nAMD [15]. Multiple different anti-VEGF agents are used for the treatment of nAMD. These therapies target one or more of the different VEGF subtypes or isoforms [16], and they may differ in their fluid-resolving properties. As examples, in the phase 3 HAWK and HARRIER studies, brolucizumab had superior fluid resolution and a longer duration of action compared with aflibercept [17], whereas generative artificial intelligence (AI) models have predicted superior fluid resolution with aflibercept versus ranibizumab [18].

Despite robust efficacy, fixed monthly injections of anti-VEGF therapies, as used in initial pivotal trials for ranibizumab [19–21], have the potential to place considerable burden on patients and healthcare systems [15, 22, 23]. Furthermore, repeated intravitreal anti-VEGF injections may be associated with potentially detrimental complications, supporting a need to clarify when anti-VEGF therapy is required [24, 25]. Accordingly, alternative dosing approaches, such as pro re nata or treat-and-extend, have been developed and widely adopted [21, 26]. Central to these alternative regimens is the ability to determine whether disease is active or quiescent [21, 26, 27]. It is therefore important to identify biomarkers of disease activity that can be used to guide treatment decisions.

IRF and SRF, as visualized by optical coherence tomography (OCT), are important biomarkers of disease activity in nAMD. Different retinal fluid types display specific patterns of distribution and SRF and IRF are of particular interest in nAMD [28–31] (Fig. 1). Between 80 and 90% of patients with nAMD have SRF [28, 32, 33], defined on OCT scan as the dark, nonreflective space between the neurosensory retina and the RPE [22]. SRF diffusely affects the macula in nAMD, extending throughout the entire macular region [29]. It is important to note that not all cases of SRF are due to neovascularization [34–36], as discussed later in this review in the context of differential diagnosis. There is less information on the prevalence of IRF than that of SRF, but it is estimated that 40–68% of patients with nAMD have IRF [28, 32, 33]. IRF is defined on OCT scan as minimally reflective round spaces within the neurosensory retina [37]. IRF is typically concentrated at the foveal center [29]. Although less well explored than SRF and IRF, sub-RPE fluid has been reported in 32–49% of eyes with nAMD [28, 38].

Fig. 1.

OCT scan illustrating the appearance of CNV, SRF, and IRF in nAMD, and presence of outer retinal tabulation. CNV choroidal neovascularization, IRF intraretinal fluid, nAMD neovascular age-related macular degeneration, OCT optical coherence tomography, SRF subretinal fluid

This article is based on previously conducted studies and does not contain any new data concerning human participants or animals. To identify relevant publications, unrestricted, non-systematic PubMed searches with a cut-off date of March 2024 were performed. Search strings used were: (“neovascular age-related macular degeneration” OR “nAMD” OR “wetAMD”) AND (“intraretinal fluid” OR “IRF”); (“neovascular age-related macular degeneration” OR “nAMD” OR “wetAMD”) AND (“subretinal fluid” OR “SRF”); and (“neovascular age-related macular degeneration” OR “nAMD” OR “wetAMD”) AND (“sub-retinal pigment epithelium fluid” OR “sub-RPE fluid”). Publications reviewed during the literature searches included original research articles, reviews, case reports, case series, editorials, and meta-analyses. Relevant search results were manually selected to ensure the inclusion of recent articles describing the appearance of retinal fluid or changes to retinal fluid in patients with nAMD only. Reference lists of articles identified during the searches were also checked to identify further literature of relevance to the current review. Only articles in English were considered for inclusion.

Classification of Retinal Fluid

Retinal diseases associated with VEGF-driven vascular exudation are typically treated using anti-VEGF therapies [22, 39, 40]. The presence and dynamics of retinal fluid are central to guiding anti-VEGF treatment and its frequency [30]. As anti-VEGF agents act to reduce the increase in vascular permeability caused by excess VEGF and inhibit MNV growth (VEGF was initially known as vascular permeability factor following its discovery by Dvorak and colleagues) [41], they reduce exudation and fluid accumulation [42]. The resolution of exudative retinal fluid is a key treatment goal [43, 44]. However, this mechanism acts on active MNV and exudation [42], making it essential to differentiate active from inactive nAMD before treatment, both in clinical practice and in the design of clinical trials.

SRF occurs in central serous chorioretinopathy (CSC) and liquefying adult-onset foveomacular vitelliform dystrophy can mimic SRF observed in AMD [45, 46]. Retinal cysts and subretinal pseudocysts may be misidentified as IRF [46, 47]. In the subsequent subsections, we discuss the characteristics, differential diagnosis, and prognosis of different types of retinal fluid.

Subretinal Fluid

Characteristics

SRF is an accumulation of exudative fluid between the outer border of photoreceptors and the inner border of the RPE [48]. In nAMD, SRF diffusely affects the entire macular region, with dominance in the parafoveal inferior section, defined on OCT scan as nonreflective spaces between the posterior boundary of the neurosensory retina and the RPE [29, 37].

Differential Diagnosis

There are two key challenges for differential diagnosis of SRF in nAMD – firstly, the potential for SRF in AMD to occur in the absence of neovascularization [34–36, 49]. The second is the potential for SRF to be observed due to conditions mimicking AMD [45, 46]. It is therefore important to ensure consistent nomenclature and diagnostic criteria are used.

SRF in the absence of subretinal neovascularization may occur in several situations. In eyes with serous pigment epithelial detachment (PED), SRF of non-neovascular origin may occur overlying the PED [34, 35, 49]. SRF may also accompany confluent drusen in the absence of neovascularization, potentially due to mechanical strain [34, 36]. As such conditions may fail to respond to anti-VEGF treatment [35], it is important for clinicians to be aware of these potential scenarios. Various conditions can be associated with SRF and resemble nAMD; Table 1 provides an overview of key characteristics that may aid the differential diagnosis of SRF associated with different conditions.

Table 1.

Differential diagnosis of SRF, IRF, and sub-RPE fluid

Disease	Clinical characteristics		Reference(s)
	Location	OCT features
SRF differential diagnosis
AMD SRF	Subretinal, primarily located outside the foveal center and extending throughout the entire macular region	Dark, nonreflective space between the neurosensory retina and the RPE	[29, 37]
CSC	Shallow subretinal fluid (in cases of chronic CSC)	Serous retinal detachment, with or without PED, and without intraretinal or subretinal hemorrhage; may be associated with pachychoroid	[45, 50, 51]
Adult-onset foveomacular vitelliform dystrophy	Subretinal, located above the RPE	Hyporeflective region can often be seen above the bright band of the RPE	[46]
IRF differential diagnosis
AMD IRF	Within the retinal layers	Minimally reflective round spaces within the neurosensory retina	[29, 37]
Retinal cysts	Within the retinal layers	Hyporeflective round or ovoid cystic structures with a hyperreflective border	[124]
Subretinal pseudocysts	Transient, usually either at the immediate border of a detached neurosensory retina or within the subretinal space as floating structures	Subretinal cystoid spaces with hyperreflective borders, at least one of which is concave or straight	[74]
Retinal tubulation	Outer nuclear/photoreceptor layer	Circular spaces with hyperreflective borders	[78]
Sub-RPE fluid differential diagnosis
Sub-RPE fluid/fibrovascular PED	Sub-RPE space	Irregular focal elevations of the reflective RPE band with an interior that is not optically empty or uniform, including hyper- and hyporeflective layers	[88]
Serous PED	Sub-RPE space	Dome-shaped elevations of the RPE with sharp borders over a hyporeflective space, with the Bruch’s membrane visible as a straight, thin, hyperreflective line at the base of the elevation	[88]
Drusenoid PED	Sub-RPE space	RPE elevations over homogeneous, mildly hyperreflective spaces, often with the Bruch’s membrane visible at the base	[88]

AMD age-related macular degeneration, CSC central serous chorioretinopathy, IRF intraretinal fluid, OCT optical coherence tomography, PED pigment epithelium detachment, RPE retinal pigment epithelium, SRF subretinal fluid

Of these, CSC is potentially the most important condition, with greatest risk of being mistaken for nAMD [45]. CSC is a relatively benign condition characterized by serous retinal detachment, with or without serous PED. CSC is associated in most patients with pachychoroid, in which the choroid is thickened (> 395 microns), accompanied by dilated Haller vessels (termed pachyvessels), and hyperpermeable [45, 50–52]. Neovascularization in association with CSC can be particularly challenging. The presence of a flat irregular PED should raise suspicion of the presence of choroidal neovascularization (CNV; in approximately 20% of cases) [53, 54]. Neovascularization is often underdiagnosed in this situation. In elderly populations, CSC may be a risk factor for the development of CNV. One study found that CNV occurred in 44.4% (n = 11) of patients with CSC aged between 65 and 91 years [45]. Patients with CSC and CNV may respond to anti-VEGF treatment with resolution of the SRF [45]; however, the symptomatic treatment of CNV does not address the underlying cause of CSC [45].

OCT angiography may be helpful for diagnosis and follow-up to facilitate differentiation of retinal diseases such as CSC from SRF related to nAMD. In CSC, OCT angiography can detect areas with high signal intensity and abnormal Haller vessel dilation, as well as reveal abnormal blood flow patterns surrounded by hyperperfusion, which is indicative of focal choroidal ischemia [55]. Dark areas, dark spots, and abnormal choroidal vascularization are typical OCT angiography imaging findings in CSC eyes. These dark areas corresponding to SRF present as focal or diffuse, ill-defined, low-detectable flow areas [55].

Beyond CSC, various macular dystrophies can resemble AMD, including adult-onset foveomacular vitelliform dystrophy [46]. Differential diagnosis can be particularly challenging in early stages of the disease, when vitelliform deposits may have a clinical appearance similar to drusen in AMD [46]. Vitelliform deposits are subretinal, and the bright band of the RPE can be observed below the vitelliform lesion [46]. A hyporeflective region is often observed and may be mistaken for nAMD [46]. OCT angiography can demonstrate a vascular network below the lesion if present, enabling differentiation between vitelliform dystrophy and exudative AMD [46], and thus SRF.

More recent approaches to detecting SRF have utilized AI-based methods in combination with OCT. These methods have several advantages over manual approaches to diagnosing SRF by allowing for fully automated and high-precision objective identification, localization, and quantification of fluid [56–58]. Additionally, AI methods can facilitate this with superior speed and accuracy versus human investigators, even those with a very high level of expertise and experience [56–58]. The increasing use of these approaches will enhance both the identification of SRF and monitoring of disease progression and responses to treatment [56–58].

Prognosis

Determining the impact of fluid in different retinal compartments on prognosis is essential to optimize outcomes and reduce the treatment burden in patients with nAMD [59]. Anti-VEGF agents have long been the standard of care in nAMD. However, in studies performed before the generalized use of anti-VEGF therapy, patients with nAMD generally typically experienced significant visual loss associated with nAMD in the first 2 years following diagnosis [60]. More recent studies in patients treated with anti-VEGF therapy have shown a dramatic improvement in prognosis. For example, the SEVEN-UP multicenter cohort study shows good visual outcomes in many patients at 7 years after treatment initiation with anti-VEGF agents. These observations indicate that treatment that reduces or resolves retinal fluid has improved visual outcomes in nAMD [61]. However, in this section, we review evidence obtained primarily from post hoc analyses of registrational trials of anti-VEGF therapies. These studies have shown that SRF in nAMD may be associated with improved visual outcomes and a reduced risk of the development and progression of macular atrophy and fibrosis [62, 63]. The potential implications of these findings for treatment decisions in clinical practice are discussed later in the article.

An analysis of the EXCITE trial in 319 patients with nAMD demonstrated that eyes with SRF at baseline had favorable visual acuity (VA) outcomes, regardless of whether patients received a monthly or quarterly anti-VEGF therapy regimen [62]. Similarly, in an analysis of the VIEW trials (N = 1815 patients with nAMD), eyes with SRF at baseline demonstrated less deterioration in VA after 96 weeks of anti-VEGF therapy versus those without [64]. In the VIEW 2 trial, a smaller reduction in best-corrected VA (BCVA) was reported in eyes with SRF than with other fluid types following 12 weeks of anti-VEGF therapy [63]. However, an analysis of the HARBOR trial in patients with nAMD receiving anti-VEGF therapy found no significant difference in VA outcomes over a 24-month period for eyes with differing SRF thicknesses [32].

The HARBOR trial showed that the presence of SRF at baseline was associated with a lower risk of atrophy over 24 months compared with eyes without SRF; this was consistent with findings from the IVAN and CATT trials [43]. In an analysis of data from 594 patients with nAMD in the IVAN trial, the presence of SRF at the final visit halved the probability of atrophy developing within an nAMD lesion (odds ratio [OR] 0.41; P = 0.004) [65]. Secondary analyses of the CATT trial in patients with nAMD also showed that a greater baseline SRF thickness (with a threshold of > 25 μm identified by one group of investigators) on OCT scan was associated with a lower risk of atrophy development [66, 67]. Similar results were seen in an observational study of 973 treatment-naïve eyes over the first 5 years of treatment. The findings demonstrated that eyes with SRF and no IRF had the lowest risk of developing subfoveal or extrafoveal atrophy (P = 0.024 and P < 0.001, respectively) and improved 5-year VA (P < 0.001) compared with eyes that had IRF or a combination of SRF and IRF [68]. Another real-world analysis of 359 patients with nAMD receiving long-term anti-VEGF therapy found that the presence of SRF at baseline was associated with a lower probability of developing atrophy and less frequent atrophy progression at 36 months (P = 0.002 for both) [69]. The same study also found that the presence of SRF at 36 months was associated with a lower rate of fibrosis (OR 0.49; 95% confidence interval [CI] 0.29–0.81) and fibrosis progression (OR 0.50; 95% CI 0.31–0.81) in patients with nAMD (both P = 0.005) [69]. Marginally lower rates of fibrosis were also found in eyes with SRF versus IRF in the observational study mentioned previously [68]. Therefore, it can be hypothesized that the presence of SRF in eyes with nAMD may be a protective biomarker against atrophy, fibrosis and progression of vision loss. A suggested mechanism for how SRF provides a beneficial effect in nAMD is that alteration of its composition (either innately or with treatment) in response to CNV to more closely mimic the normal extracellular milieu of photoreceptors, acting as a survival response against macular atrophy [70]. Based on the outcomes of the FLUID study, which showed that BCVA outcomes were noninferior if some residual SRF was tolerated versus total resolution [71], it is the treating clinician’s discretion as to whether the complete resolution of SRF is needed or not. These findings have potential implications for whether complete resolution of SRF is required, as discussed in the “Clinical Practice” section of this article.

Intraretinal Fluid

Characteristics

IRF can be defined as an accumulation of fluid within the inner retinal layers, typically concentrated to the foveal center, and is often associated with increased retinal thickening [29, 48]. On OCT scan, IRF is defined as round, minimally reflective spaces within the neurosensory retina [37].

Differential Diagnosis

Several conditions can resemble IRF at presentation; however, consistency in the published terminology is lacking. Table 1 provides an overview of the key clinical characteristics of various features that may resemble IRF to aid differential diagnosis. IRF may be misidentified as retinal cysts, subretinal pseudocysts, or retinal tubulation. For example, in the CATT trial, the term IRF was used to describe hyporeflective intraretinal cystoid structures but did not distinguish between structures with or without a hyperreflective border [47]. Even when the reflectivity of a cystoid space border is considered, the appearance of IRF on OCT scans closely resembles that of retinal cysts. These findings can be differentiated based on treatment response, as IRF is sensitive to anti-VEGF therapy and will resolve quickly, whereas retinal cysts will not [29, 72]. Schmidt-Erfurth et al. termed both IRF and retinal cysts as forms of intraretinal cyst, dividing these into exudative and degenerative categories [63]. Exudative cysts were defined by their quick resolution after an initial loading dose of intravitreal anti-VEGF therapy (IRF), whereas degenerative cysts were defined by their persistence (intraretinal cysts), with the latter being termed “pseudocysts” [63, 73]. It is also established that intraretinal cysts in untreated nAMD are the only feature that correlates with VA and are a predictor of poorer visual outcomes [63, 64]. Therefore, it is key to differentiate between IRF and retinal cysts, and implement consistent and robust definitions for them, especially given this difference in treatment response.

Subretinal pseudocysts are defined as optically empty subretinal cystoid-like spaces with hyperreflective borders and at least one concave or straight border. They occur at the immediate border of the detached neurosensory retina or within the subretinal space as floating structures and are likely transient alterations of the photoreceptor outer segments and RPE [74, 75]. A key differentiator for pseudocysts is that no retinal thickening occurs in the involved retinal area compared with the adjacent retina, in contrast to IRF or retinal cysts [76]. Such pseudocysts are not associated with neovascularization but are common in geographic atrophy (GA) [76, 77]. Such pseudocysts can be seen in eyes with fibrotic scarring, suggesting that these changes are nonexudative in nature and would therefore not require anti-VEGF therapy for nAMD [75]. Subretinal pseudocysts may be a degenerative change that does not require therapy [75], they are associated with photoreceptor loss, incomplete definition of the RPE, signs of retinal atrophy and less improvement in vision in response to therapy [74].

IRF is also distinct from outer retinal tubulation, a common feature of nAMD characterized by the circular or tubular rearrangement of photoreceptors following retinal injury (Fig. 1) [78]. On OCT scans, this is seen as circular, cyst-like spaces in the outer nuclear layer [78], often at the border between normal and abnormal retina, with hyperreflective borders thought to be derived from the ellipsoid layer [47, 78]. Retinal tubulation most often occurs in eyes with MNV type 2 disease and underlying fibrosis, contraction of which may be a causative factor [79]. As a nonexudative feature, retinal tubulations do not respond to anti-VEGF therapy and are associated with a reduced rate of progression to GA [80, 81].

Prognosis

Unlike SRF, IRF is associated with poorer visual outcomes and an increased risk of developing atrophy and fibrosis, both of which contribute to vision loss [43, 62]. In the CATT trial of eyes with nAMD treated with anti-VEGF therapies, IRF was found to have a greater impact on VA than SRF or sub-RPE fluid at all time points [82, 83]. This was particularly apparent for IRF in the foveal center, suggesting that retinal fluid location may be important with regard to long-term impact [83]. In a retrospective analysis of 301 eyes with nAMD treated with anti-VEGF therapy over 12 months, high levels of IRF and variability in IRF were also associated with worse VA outcomes [84]. The VIEW 2 study mirrored these findings, while also showing that IRF was the only fluid type significantly associated with poor BCVA at baseline, at − 5.98 (standard error ± 0.85) letters (P < 0.0001), with a further reduction in BCVA by 3.92 (standard error ± 1.04) letters after 3 months of treatment with anti-VEGF therapy [63].

In the 5-year follow-up for the Fight Retinal Blindness registry study of 973 eligible eyes with nAMD, IRF was associated with lower VA and greater risk of developing atrophy than SRF [68]. Another real-world study of patients with nAMD receiving long-term anti-VEGF therapy found that IRF at baseline was associated with an increased risk of developing atrophy (OR 1.87; 95% CI 1.09–3.20) [69]. Likewise, IRF at baseline was reported as a risk factor for the development of macular atrophy by month 24 of anti-VEGF therapy in a post hoc analysis of the HARBOR trial [43]. This is further supported by the results of the CATT trial, which demonstrated that IRF at baseline and following 2 years of treatment with an anti-VEGF agent was a risk factor for atrophy [66]. Although a greater effect was seen from IRF at the foveal center (hazard ratio 2.10; 95% CI 1.34–3.31), IRF outside of this area was also associated with a greater risk of atrophy compared with the absence of IRF (hazard ratio 1.80; 95% CI 1.10–2.95) [66].

IRF is also a risk factor for fibrosis, with the Fight Retinal Blindness registry study reporting a higher incidence of extrafoveal fibrosis in eyes with mostly IRF than those with mostly SRF or inactive disease [68]. Furthermore, a real-world study of 359 patients with nAMD demonstrated that the presence of IRF at any visit was strongly associated with the development and progression of subretinal fibrosis [69]. Finally, an analysis of 569 treatment-naïve eyes with nAMD that were treated with a disease activity-dependent treat-and-extend-based anti-VEGF regimen found that, rather than the presence or absence of IRF, an increase in IRF volume was associated with an increased risk of developing fibrosis [85]. In light of the evidence presented here, we can state that IRF is a biomarker for poor prognosis for eyes with nAMD, and its resolution should be a key treatment goal.

Subretinal Pigment Epithelium Fluid

Characteristics

Sub-RPE fluid is an accumulation of exudate between the RPE and the inner collagenous layer of the Bruch’s membrane, causing a serous retinal PED, and occurs prior to formation of IRF in the nAMD disease course [48, 63]. On OCT scan, PED is defined by some investigators as focal elevation of the reflective RPE band over an optically clear or moderately reflective space with a minimum width of 400 µm at the base or a minimum thickness of 200 µm from the surface of the RPE band to the surface of the choriocapillaris [63, 86]. However, it should be noted that this definition is not used in all studies [87].

Differential Diagnosis

PED is a consequence of sub-RPE fluid accumulation, a common manifestation of nAMD [88]. However, PED has many causes (e.g., CSC) and is not always associated with MNV, as previously noted. Within the context of nAMD, there are three mechanisms of the formation of PED [88]:

• Serous PED: Progressive deposition of materials and thickening of the Bruch’s membrane with age, resulting in a gradual increase in fluid exudation into the sub-RPE space [88].

• Drusenoid PED: Progressive accumulation of lipoprotein particles into the sub-RPE basal space, leading to the formation of drusen and a gradual separation of the RPE from the Bruch’s membrane; not associated with underlying MNV or exudation [88–90].

• Fibrovascular/hemorrhagic PED: Formation of MNV leading to ingrowth of leaking vessels into the sub-RPE space through the Bruch’s membrane [88, 90].

Key clinical characteristics of these different forms of PED are summarized in Table 1 to aid differential diagnosis, and are further discussed below.

Serous PEDs are found in approximately 10% of eyes with AMD of any stage and are largely asymptomatic, with little or no effect on vision; however, 32–39% will progress to MNV [88]. On OCT scan, serous PEDs present as dome-shaped elevations of the RPE with sharp borders over a hyporeflective space. The Bruch’s membrane is also visible as a straight, thin, hyperreflective line at the base of the elevation [88]. There are currently no available treatments for serous avascular PED, and most cases are managed clinically through regular follow-ups and monitoring [88].

Drusenoid PEDs are a feature of intermediate, dry AMD, occurring in 8% of all cases of AMD and presenting as RPE elevations over homogeneous, mildly hyperreflective spaces, often with the Bruch’s membrane visible at the base on OCT scans [88]. The Age-Related Eye Disease Study found that, over a 5-year period, AMD had progressed from intermediate to advanced disease in 42% of eyes with a drusenoid PED, 19% had developed central-vision involving GA, and VA had decreased by ≥ 15 letters in 40% of eyes, indicating that it may be a marker of progression [91]. Currently, there are no recommendations for treatment of drusenoid PEDs, but they should be monitored for signs of progression [88].

Fibrovascular PED is the only form of PED that results from MNV, occurring in 62–80% of eyes with nAMD. This form of PED is associated with poor visual outcomes, with approximately 50% of patients experiencing significant VA loss (> 3 lines) during the first year if left untreated [88]. A fibrovascular PED appears as an elevation of the RPE with an irregular surface, often including serous and fibrous or hemorrhagic areas. In contrast to serous PEDs, fibrovascular PEDs usually appear as irregular elevations of the RPE with an interior that is not optically empty on OCT scan [88]. A key differentiator from drusenoid PEDs is the optically nonuniform interior with hyper- and hyporeflective layers [88]. Unlike other forms of PED, fibrovascular PEDs can be treated with anti-VEGF therapies due to their exudative nature. However, it is important to note that, even with appropriate treatment, eyes with fibrovascular PED are associated with worse visual outcomes and treatment resistance than eyes without fibrovascular PED [88].

Prognosis

The relationship between sub-RPE fluid and visual prognosis in nAMD is less obvious than that of SRF or IRF. For example, in the CATT trial, the presence of sub-RPE fluid, whether foveal or nonfoveal, had a significant negative impact on the proportion of patients gaining three or more lines of VA at 52 weeks with anti-VEGF therapy (P = 0.01). However, when controlling for foveal involvement of each fluid type, total fluid thickness and follow-up duration, VA was not associated with the presence of sub-RPE fluid [92]. A further analysis of the CATT trial found that sub-RPE fluid was protective against the development of atrophy [67].

When accumulation of fluid in the sub-RPE space progresses, it leads to PED, a negative indicator of disease activity. However, the VIEW 2 study found that low baseline BCVA was accounted for by the presence of IRF and that the presence or absence of PED had a comparatively limited effect on BCVA [63]. While low-level PEDs may have a limited positive effect on prognosis, those of greater than > 400 µm, with a small MNV to PED size ratio, the presence of radial hyperreflective lines in PED lesions or micro-rips in the RPE are at risk of RPE tears, and serous PEDs are at greatest risk [93]. RPE tears are serious vision limiting events in most patients. While PEDs are a major risk factor for RPE tears, there is also evidence that anti-VEGF injection may also increase the risk of tears to the RPE [93].

Combination of Features

In practice, each of the discussed retinal fluid types do not only occur in isolation; the implications of combinations of retinal fluid features must therefore also be considered. A real-world study of patients with nAMD undergoing long-term treatment found that eyes with both SRF and IRF had similar VA to eyes with IRF alone following the loading phase of anti-VEGF treatment. This study also found that fibrosis and functional progression were more frequent in patients with both SRF and IRF than those with SRF alone, indicating that outcomes are dictated by their most adverse component [69]. Further to this, in a post hoc analysis of the HARBOR trial, improvements in VA over 24 months were the greatest in eyes with residual SRF and resolved IRF, followed by eyes with resolved SRF and IRF. Eyes with residual SRF and IRF, and eyes with residual IRF and resolved SRF had the worst outcomes [32, 94]. This evidence indicates that IRF is the key determinant of prognosis, and, SRF appears to have little independent significance for visual outcomes and may even be protective.

Associations of Retinal Fluid and Macular Neovascularization

Neovascularization of the macula is a primary pathological mechanism of nAMD, leading to the exudation of fluid. MNV is classified into three subtypes, the assessment of which provides information on disease prognosis and can help determine the optimal treatment regimen [48, 95]. Although there is heterogeneity in the presentation and underlying pathology of nAMD, each of the three primary types of MNV may lead to exudation and retinal fluid accumulation with varying distribution and extent (Fig. 2; Table 2) [11].

Fig. 2.

The characteristics and treatment of MNV types. The characteristics, fluid distribution, and treatment recommendations for MNV type 1 [48, 95, 96, 123], type 2 [48, 95], and type 3  [48, 123]. *Due to the presence of large, mature vessels in the neovasculature, which may produce persistent SRF or SRF with inadequate response. IRF intraretinal fluid, MNV macular neovascularization, RPE retinal pigment epithelium, SRF subretinal fluid, VEGF vascular endothelial growth factor

Table 2.

Outcomes in nAMD by MNV type and retinal fluid compartment

MNV type	SRF		IRF
	Frequency	Outcome	Frequency	Outcome
MNV type 1	Common [69, 123]	Lower risk of progression of atrophy and slower growth rate than other subtypes [69]	May present with RPE and outer retinal atrophy in the absence of SRF [123]	Associated with significant worsening of visual acuity [125]
MNV type 2	Less common than type 1 [69]	NR	Common [69]	Increases the risk of developing fibrosis, worsening vision [69]
MNV type 3	Can occur, usually in conjunction with IRF [95]	Numerically greater incidence of macular atrophy than IRF alone [126]	Common [123]; present in nearly every case [127]	May be associated with good visual acuity and less subfoveal atrophy than MNV type 3 with SRF and IRF [128]

The three MNV types describe distinct phenotypes within nAMD. Although retinal fluid location is not the only factor considered in MNV classification, outcomes for eyes with each MNV type correspond to the fluid distribution most commonly associated with said subtype, indicating that SRF and IRF may be significant prognostic factors. Here we show that MNV type 1 is most associated with SRF and has generally better outcomes, whereas MNV type 3, which almost always features IRF, is associated with worse outcomes

IRF intraretinal fluid, MNV macular neovascularization, nAMD neovascular age-related macular degeneration, NR not reported, RPE retinal pigment epithelium, SRF subretinal fluid

MNV type 1, or ‘occult’ neovascularization, is characterized by the presence of neovascularization beneath the RPE layer and often causes SRF to accumulate. Neovascularization of choroidal origin in the neuroretina, having broken through the RPE into the subretinal space, is characteristic of MNV type 2. MNV type 3 is characterized by anomalous vascular complexes originating from the deep retinal capillary plexus in the inner retina [95]. Retinal fluid distribution varies between these subtypes, with SRF being the most common form of fluid accumulation in MNV type 1, although it may occur in MNV types 2 and 3. IRF frequently occurs in MNV types 2 and 3, but it may be observed in MNV type 1. In MNV type 3, IRF is the most common form of retinal fluid accumulation, but SRF may occur in conjunction with IRF [95].

Assessment of MNV type in nAMD provides information on disease prognosis and can help with determining the optimal treatment regimen [95]. MNV type 1 has been reported to require more frequent anti-VEGF injections than other subtypes, potentially due to the presence of large, mature vessels in the neovasculature leading to refractory or persistent SRF [95, 96]. However, MNV type 1 is also associated with more favorable long-term outcomes than other MNV types [95, 96], as type 1 is most commonly characterized by SRF [48]. This finding is consistent with the potential protective role of SRF discussed previously, and the tendency to overtreat, supporting a need for pragmatic treatment regimens that minimize the burden of treatment on such patients. MNV type 2 typically responds rapidly to treatment but may be associated with the development of fibrotic scars, leading to poorer outcomes [95]. MNV type 3 usually responds well to anti-VEGF therapy but may be associated with a higher incidence of GA than other subtypes [95].

Although MNV types are characterized by more than just the location of retinal fluid, these classifications corroborate the theory that SRF may be protective whereas IRF is more likely to negatively affect visual prognosis. However, retinal fluid location should not be considered in isolation when discussing fibrosis and atrophy. In addition to the impact of SRF and IRF, risk factors for fibrosis include MNV and MNV lesion size, hemorrhage, and subretinal hyperreflective material (SHRM) [97].

Location of Subretinal Fluid Versus Intraretinal Fluid and Associated Physiological Impact

The general location of SRF and IRF may explain their relatively different effects on vision. A post hoc analysis of the HARBOR trial showed that foveal SRF and IRF were more strongly correlated with VA than parafoveal fluid in patients with nAMD [98]. However, a separate subanalysis of the same trial found that IRF was associated with worse visual outcomes than SRF, regardless of severity or location [32].

In nAMD, SRF, in the presence of IRF, may be an indicator of milder or more benign disease, and may contain neuroprotective factors that promote RPE and outer retina survival [48]. Although SRF increases the diffusion distance between the RPE and photoreceptors, it appears that its impact on the visual cycle is relatively minor. It is hypothesized that foveal SRF may protect photoreceptors from damage, leading to improved VA [44, 48]. However, SRF occurs most commonly in MNV type 1, the stage of neovascularization that is associated with low risk of progression and a slow rate of lesion growth, and it may be coincidental that eyes with SRF have improved prognosis [48, 99]. The level of SRF is also a key factor when considering physiological impact. This is because evidence suggests that up to 150 µm thickness of SRF is tolerated with minimal negative impact on vision, but that a thickness of > 200 µm begins to impact vision [100]. In the absence of atrophy and SHRM, the effect on vision increases with thickening SRF, but if atrophy and/or SHRM are present, the impact of thickening SRF is reduced. It has been suggested that SRF associated with atrophy and SHRM may be caused by RPE pump failure, and therefore, not directly related to vision loss, or that alongside the negative effects of atrophy and SHRM, the impact of SRF on vision is comparatively minor [100].

The Clinical Implications of Subretinal Fluid and Intraretinal Fluid

Clinical Trial Design

Resolution of fluid is often used as an endpoint in clinical trials of retinal diseases as a biomarker of disease activity in nAMD [28, 29]. However, retinal fluid resolution may not always translate into effects on VA. It is therefore important to correlate the anatomical change of retinal fluid resolution with self-reported improvement in visual function to ensure that structural change reflects symptomatic improvement. Accurately assessing treatment outcomes may also reduce the need for invasive and burdensome treatment [101]. This is because subsets of patients with certain baseline characteristics in combination with SRF or IRF may be preferentially selected for more ‘relaxed’ treatment regimens such as that used in the FLUID trial, as discussed in more detail below [71]. Recent studies including TENAYA and LUCERNE, and PULSAR have investigated the alternative physiological endpoint of change in central subfield thickness, and allowed for the retention of some residual fluid [102–104]. However, in these eyes atrophy would cause retinal thinning, thereby masking any thickening from IRF accumulation. Although such alternative measures may be clinically relevant, clinical trials in nAMD require robust endpoints to assess disease activity and prognostic markers. The inclusion of patient-reported outcome measures and identification of biomarkers that may predict stability of retinal fluid are vital to reduce the treatment burden for patients whose disease is unlikely to worsen by tolerating SRF. They may also allow for earlier patient stratification within the treatment paradigm.

A further aspect of clinical trials in which fluid status data could be better utilized is stratification of treatment arms. As retinal fluid distribution is a prognostic factor that could influence treatment decisions [105], an uneven distribution of fluid status between treatment arms could potentially introduce bias. While fluid status has been used in some studies and post hoc analyses, the majority of clinical trials in nAMD at present rely on randomization or employ minimal stratification factors [32, 63, 106, 107].

Clinical Practice

Evidence suggests that the relationship between retinal fluid and visual outcomes is complex and that while the aim of treatment should be to resolve IRF, complete SRF resolution may not always be necessary [32]. Satisfactory anatomical and functional outcomes may be achieved by tolerating some residual SRF, potentially enabling patients with nAMD to be treated using fewer injections—the 24-month FLUID trial provides important insights in this regard. An aim of FLUID was to establish whether SRF requires complete resolution in patients with nAMD [71]. Participants were randomized to receive ranibizumab monthly using either an ‘intensive’ regimen requiring complete resolution of IRF and SRF prior to extending the treatment interval or a ‘relaxed’ regimen requiring complete resolution of IRF but tolerating residual SRF up to 200 µm at the foveal center [71]. Visual outcomes were similar between the intensive and relaxed regimens, indicating that treatment intervals can be extended with some residual SRF present [71]. In addition, a retrospective analysis of 50 patients with nAMD receiving aflibercept demonstrated that a ‘relaxed’ regimen yielded similar efficacy and safety outcomes at 12 months versus an ‘intensive’ regimen [33]. One model suggests that SRF is well tolerated if the external limiting membrane is intact, and much less so if not [108], and that SRF of up to 150 µm has minimal impact on vision [100].

The impact of variations in retinal fluids over time may also influence outcomes, although this is not yet well understood. A post hoc analysis of the phase 2 OSPREY trial from Ehlers et al. [109], found that greater SRF volatility was associated with a greater increase in SHRM and lesser improvement in ellipsoid zone integrity. These are both negative predictors of visual outcome, suggesting that variability is detrimental to anatomic features in nAMD [109].

In contrast, post hoc analysis of the CATT trial reported that in patients with nAMD, subretinal fluid was significantly associated with a decreased risk of GA progression (P < 0.001) over a 2-year follow-up [61]. Furthermore, post hoc analysis of functional outcomes of the HARBOR study demonstrated that foveal SRF, but not parafoveal SRF or IRF correlated with better BCVA at baseline and subsequent visits, and that after the first injection, SRF reductions were associated with visual gain [98]. Therefore, there may be a need to balance visual outcomes with the burden of regular, intensive treatment.

Patient preferences should also be considered in treatment decision-making. In studies of patients receiving treatment for nAMD, achieving and maintaining good vision was ranked as the most important factor in their treatment. However, other factors, including having the assessments and treatments at the same time and location, less frequent follow-up visits, and reduced waiting times were also ranked highly by patients [110]. This finding, alongside those of other patient preference studies, adds credence to the idea that SRF may not need to be fully resolved in all cases for optimal patient outcomes [110–112].

This poses the question of which patients with nAMD may tolerate some residual SRF without clinical worsening and, therefore, could be treated with fewer anti-VEGF injections. Identifying the type of MNV may be important when determining which patients may tolerate some SRF. Not all SRF caused by active MNV will indicate a need for anti-VEGF therapy [33, 71]. Other factors including SHRM, fibrosis and atrophy have been associated with a poor response to treatment [113–115] and should also be assessed when determining appropriate treatment [107]. These are predominantly features of permanent structural damage rather than biomarkers of disease activity [11, 43]. Therefore, resolving SRF will not necessarily improve vision as the issue of permanent structural damage will remain unresolved.

In contrast to SRF, IRF is detrimental to vision and should be treated [62, 82]. However, other conditions can mimic the presentation of IRF, for instance, cyst-like outer retinal tubulations in the outer retina. These tubulations are a common feature of advanced AMD, in which photoreceptors form circular or ovoid lesions as they degenerate [80]. Tubulations will not be resolved by anti-VEGF therapy [116]. Likewise, retinal pseudocysts may be mistaken for IRF [73]; anti-VEGF agents do not affect such pseudocysts owing to their nonexudative nature [73]. Identification of cysts in these patients is important as overtreatment with anti-VEGF therapy could be detrimental.

It has been hypothesized that intensive, long-term anti-VEGF therapy may exacerbate degeneration in the retina, potentially leading to earlier progression to GA. Evidence suggests that the choriocapillaris is dependent on VEGF support from the RPE [117–119]. Consequently, continual anti-VEGF therapy would be expected to increase the risk of GA [118, 119]. Indeed, in the CATT trial, there was a 59% increase in GA over 2 years with monthly anti-VEGF injections compared with pro re nata treatment, regardless of the anti-VEGF agent used. In most studies, patients are followed for up to 2 years; however, the natural turnover of retinal endothelial cells can take more than 2 years [66, 120, 121]. Considering this, longer-term studies of anti-VEGF therapy outcomes are needed to assess the true impact of these injections.

Identification of new OCT imaging biomarkers for nAMD, including those based on retinal fluid distribution, may enable personalized treatment tailored to each patient’s individual disease characteristics. It may also allow for enhanced monitoring of the disease course, thereby reducing the treatment burden [31, 48].

Conclusions

SRF and IRF remain important biomarkers of disease activity in nAMD, with retinal fluid resolution representing a key goal of anti-VEGF therapy and a key clinical trial endpoint. Resolution of IRF should be targeted as it is associated with worse visual outcomes and an increased risk of developing atrophy and fibrosis. However, a growing body of evidence has shown that SRF may be associated with improved visual outcomes, reducing the risk of development and progression of macular atrophy and fibrosis. Our key clinical recommendations regarding SRF and IRF in nAMD can be found in Fig. 3.

Fig. 3.

Proposed algorithm for diagnosis, prognosis, and treatment of SRF and IRF in nAMD. IRF intraretinal fluid, MNV macular neovascularization, nAMD neovascular age-related macular degeneration, RPE retinal pigment epithelium, SRF subretinal fluid, VEGF vascular endothelial growth factor

Definitively determining the impact of retinal fluid in different retinal compartments is essential to optimize outcomes and reduce the treatment burden in patients with nAMD. As some residual SRF may be protective of vision, the treatment strategy could potentially be altered in some cases. The goal of this would be to reduce treatment burden and prevent the potentially negative adverse events associated with intravitreal administration of anti-VEGF therapy, thereby reducing associated healthcare costs. However, because some studies have shown that complete retinal fluid resolution can lead to even greater benefit, at present, endorsing a potential protective effect of mild SRF cannot be included in clinical recommendations.

Retinal fluid location should not be considered in isolation when discussing fibrosis and atrophy. In addition to the impacts of SRF and IRF, risk factors for fibrosis include MNV and MNV lesion size, hemorrhage, and SHRM. It is important to remember that there is a trade-off here, and that retinal fluid should always be treated to some extent, if possible. However, SRF may be tolerated if it cannot be resolved with a reasonable treatment regimen. The aforementioned factors should be considered in cases of persistent SRF and may be used to guide treatment decisions. However, cause and effect between SRF and reduced or slowed atrophy has not yet been proven, and further research is required to establish this.

Examples of future studies that might be conducted to fill the literature gaps identified within this review may include a randomized trial comparing the visual function benefits of residual SRF versus total SRF resolution. Studies to more clearly establish the relationship between sub-RPE fluid and visual prognosis would also be of interest, as would assessments of longer-term anti-VEGF therapy outcomes, identification of new OCT imaging biomarkers for nAMD based on retinal fluid distribution, and studies of the impact of variations in retinal fluids over time on disease and treatment outcomes.

With a well-defined algorithm aligning to the recommendations given in this paper, we may be able to optimize outcomes of patients with different types of retinal fluid. In the future, automated, artificial intelligence-based algorithms may aid with identifying and quantifying different retinal fluid types and help guide treatment decisions, but require careful validation [30, 56, 122].

Author Contributions

All authors meet the criteria for authorship as recommended by the International Committee of Medical Journal Editors (ICMJE). The article concept was conceived by the authors. The literature search was performed as part of the medical writing support. Alexander JE Foss, David Almeida, Chui Ming Gemmy Cheung, Yuichiro Ogura, Eduard de Cock, and Theo Empeslidis contributed to the drafting, review, and revision of the manuscript, and approved the final manuscript.

Funding

No funding or sponsorship was received for this article. All fees for publication of the article were funded by Boehringer Ingelheim.

Data Availability

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.

Declarations

Conflict of Interest

The authors did not receive payment related to the development of this manuscript. Boehringer Ingelheim was given the opportunity to review the manuscript for medical and scientific accuracy, as well as intellectual property considerations. Alexander JE Foss has received writing support from Boehringer Ingelheim. David Almeida is Co-Founder and Chief Medical Advisor of Citrus Therapeutics and holds equity with this company. David Almeida declares involvement with AbbVie, ACELYRIN, Alcon, Alimera Sciences, Allergan, Bausch + Lomb, Bayer, Boehringer Ingelheim, Dutch Ophthalmics, EyePoint Pharmaceuticals, Genentech, Gyroscope Therapeutics, Novartis, Opthea, Regeneron, REGENXBIO, Roche, Samsara Vision and Spherix Consulting Group. Chui Ming Gemmy Cheung declares involvement with Bayer, Boehringer Ingelheim, Iveric Bio, Novartis, Roche, Topcon, and Zeiss. Yuichiro Ogura declares consultancy for Alcon Japan, Apellis, Bayer, Boehringer Ingelheim, Chugai Pharmaceutical Co., Ltd, HOYA, Iveric Bio, Kyoto Drug Discovery & Development, Novartis, Senju; research support from Bayer, Kowa, Nikon Healthcare, Novartis, Santen, Topcon, and ZEISS; and is an employee of Genentech, Inc. Eduard de Cock and Theo Empeslidis are employees of Boehringer Ingelheim.

Ethical Approval

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.