Recent advancements in the management of retinoblastoma and uveal melanoma

Retinoblastoma in children and uveal melanoma in adults can pose a serious threat to both vision and life. For many decades, enucleation was often the only option to treat these intraocular malignancies. For retinoblastoma, intra-arterial chemotherapy is often utilized as the primary treatment at advanced academic centers and has dramatically improved local tumor control and eye salvage rates. For uveal melanoma, both plaque brachytherapy and proton beam irradiation have served as widely utilized therapies with a local failure rate of approximately 1–10%, depending on the series. Major recent advancements have allowed for a better understanding of the genomics of uveal melanoma and the impact of certain mutations on metastatic susceptibility. Gene expression profile stratifies uveal melanomas into two classes: low-risk (class 1) and high-risk (class 2). A loss-of-function mutation of BAP1 is associated with a class 2 gene expression profile and therefore confers worse prognosis due to elevated risk of metastasis. On the other hand, gain-of-function mutations of EIF1AX and SF3B1 correspond to a gene expression profile of class 1A and class 1B and confer a better prognosis. Preferentially expressed antigen in melanoma (PRAME) is an antigen that increases metastatic susceptibility when expressed in uveal melanoma cells. In addition to plaque brachytherapy and proton beam irradiation, both of which have demonstrated superb clinical outcomes, scientists are actively investigating newer therapeutic modalities as either primary therapy or adjuvant treatment, including a novel nanoparticle therapy and immunotherapy.


Introduction
Eye cancer is a rare but potentially life-threatening disease. Although there are various kinds of intraocular malignancies, retinoblastoma and uveal melanoma are the most common primary intraocular malignancies. Retinoblastoma primarily occurs in children, with 250-350 cases per year in the United States almost exclusively in patients under 5 years of age 1 . Uveal melanoma is the most frequently observed primary intraocular cancer in adults, with 5.1 cases per million mostly seen in Caucasians 2 . In this review, we will provide updates on treatment guidelines, newly emerging therapeutic modalities, areas of research focus, and future directions for the management of retinoblastoma and uveal melanoma.

Retinoblastoma
Retinoblastoma is rare in the general population but is the most common primary pediatric intraocular malignancy. Retinoblastoma occurs when both alleles of the RB1 gene, which is a tumor suppressor gene located at chromosome 13q14, are mutated. Approximately one-third to 40% of retinoblastoma is attributable to a mosaic or germline mutation of the RB1 gene 3 , while sporadic cases account for more than 50% (Table 1). Almost 90% of the germline cases are due to a new mutation with no familial history of retinoblastoma 3 . Retinoblastoma can present as either unilateral or bilateral disease, which accounts for approximately two-thirds and one-third of all cases, respectively. More than 80% of unilateral retinoblastoma arises from a sporadic mutation and 15% due to a germline mutation 4 . Less than 2% of unilateral, non-familial retinoblastoma patients may present with fully functional RB1 but instead with a MYCN oncogene mutation. Patients with MYCN mutations present early at a median age of 4 months with an aggressive course 5 .
In contrast to unilateral cases, bilateral retinoblastoma is exclusively due to a germline mutation, which may be present already in the germ cell or can occur de novo during early embryogenesis. In the former, every cell in the body will carry the mutation, whereas the latter will cause mosaicism in which a proportion of cells are mutated. In either case, there is increased risk of malignancy in other parts of the body during the lifetime of the patient, including soft tissue sarcomas, osteosarcoma, and cancers of epithelial origin 6,7 . It is believed that germline retinoblastoma can also lead to an approximately 5% risk of an intracranial malignancy, most commonly pineoblastoma. Many centers perform routine monitoring with imaging even after successful intraocular treatment at least up until 5 years of age to evaluate for this second malignancy 8 . Recent controversial evidence suggests that secondary central nervous system (CNS) tumors occur in only about 5% of patients with germline mutations, especially if the patient did not receive any form of radiotherapy. Some centers have transitioned away from routine CNS surveillance after retinoblastoma treatment because of a low yield of detecting CNS tumors post-treatment 9 . However, it should be noted that pineoblastoma and trilateral retinoblastoma, which are of particular concern for germline retinoblastoma patients, were excluded from analysis. Therefore, without a large prospective study, it is premature to recommend against no routine CNS surveillance after standard retinoblastoma treatment.
With major advancements in retinoblastoma treatment in the past two decades including intra-arterial chemotherapy (IAC) and intravitreal chemotherapy (IvitC), advanced academic centers across the globe have achieved over 95% long-term disease-free survival and 85% long-term globe salvage, which will be discussed further below 10 .

Retinoblastoma classification and current mainstay treatment options
Both Reese-Ellsworth and the International Classification of Retinoblastoma (ICRB) systems can be used to classify retinoblastoma (Table 2). However, the Reese-Ellsworth classification was introduced decades ago when external beam radiation therapy (EBRT) was primarily used, and the ICRB system (in which tumors are grouped into A through E) has largely replaced its older counterpart in current ocular oncology practice 11 . Ideally, the goal of retinoblastoma treatment is to cure the patient and secondarily preserve the globe and vision when possible. For ICRB group A, B, and some C tumors, focal therapy with laser ablation, cryotherapy, or transpupillary thermotherapy (TTT) may be sufficient to achieve full local tumor control 12 . On the other hand, group C tumors involving the macula or group D or E need to be more aggressively treated. Intravenous chemotherapy (IVC) using vincristine, carboplatin, and etoposide has been successfully used for decades to reduce the size of the tumor followed by focal therapies, with its efficacy demonstrated in multiple publications [13][14][15] . IVC can achieve over 90% of tumor control for group A-C tumors and is also frequently used to treat group D retinoblastoma, albeit with lower ocular salvage rates 15,16 . Group D or E tumors with high-risk features, including concern for extraocular extension  18 . Notably, less than 2% of the patients experienced metastasis and 1% died at 20 years of follow-up, demonstrating excellent long-term disease control. In addition to advanced unilateral disease, IVC can also be used to manage bilateral retinoblastoma with a germline RB1 mutation.
Since the early 2000s, IAC has emerged as a promising alternative for IVC and enucleation ( Figure 1). Not only is IAC a globe-salvaging treatment, but it has also demonstrated superior local tumor control with significantly fewer systemic toxicities, including immunosuppression, ototoxicity, and nephrotoxicity, than IVC 16,20-28 . In particular, for group D tumors, which had been conventionally treated with IVC or enucleation, numerous studies using IAC as primary therapy have reported 85-90% globe salvage rates 16,21,29,30 . Therefore, large academic centers with experienced ocular oncologists and neuro-interventionalists now prefer to utilize IAC as first-line treatment for most cases of retinoblastoma. Moreover, IAC has demonstrated efficacy as secondary treatment. Over 80% and 50% ocular event-free survival has been observed for patients who underwent IAC as secondary therapy after failing IAC 31 or IVC and EBRT 32 , respectively.
For very young patients, usually under 6 months of age, either bridge therapy or IAC is currently used. In bridge therapy, IVC is performed at the time of diagnosis for initial disease control until the patient reaches a weight at which IAC can be safely administered (typically 6 to 10 kg) 33 . This convention is to minimize procedure-related complications in small infants, such as catheterization site hematoma and limb ischemia. However, more studies are demonstrating that IAC can be safely administered in infants weighing less than 10 kg, with the youngest reported patient only 35 days of age when receiving the first cycle of IAC 34-38 . Sweid et al. published a comparative study in which patients weighing less than 10 kg experienced fewer total IAC cycles as well as less exposure to radiation during fluoroscopy, which can help minimize procedure-related adverse effects in the long term 34 . As a result, there is increasing understanding that younger patients undergoing IAC as primary therapy may attain better outcomes in a shorter treatment duration 29,34,39 .
For ICRB groups C, D, or E retinoblastoma with recurrent or persistent vitreous seeds after primary therapy (either IVC or IAC), IvitC is often utilized 40,41 . The technique is performed with measures to minimize needle tract tumor seeding after injection and is now widely used for various types of vitreous seeds. Previously, vitreous seeds were considered higher risk and often warranted enucleation because both IAC and IVC had limited bioavailability within the vitreous and therefore had limited efficacy. In contrast, IvitC allows for direct injection of the chemotherapeutic agent into the intraocular space and has achieved excellent seed control in multiple studies [42][43][44] . It has therefore reduced the rate of enucleation due to persistent vitreous seeds in many cases. Rates of ocular salvage in many recent publications reflect the combination use of IvitC and IAC and thus may not reflect results that would be obtained with IAC alone. Melphalan is the most commonly used agent and usually achieves excellent vitreous seed control alone. However, if there is concern for resistance to melphalan, topotecan may be added simultaneously or separately to augment intravitreal therapy 17 . In some instances, seeds from the primary retinoblastoma lesion may disseminate to the anterior chamber. Previously, enucleation was usually warranted when tumor seeding in the aqueous humor was observed.
There are reports of successfully controlling aqueous seeds with intracameral injection of melphalan, but larger studies are needed to further validate the efficacy of intracameral chemotherapy [45][46][47] .
Factors that affect primary treatment choices and treatment outcomes There are several variants of the ICRB that are currently used, including the Children's Oncology Group Philadelphia 48 and Children's Hospital of Los Angeles versions 11 . The major implication of using different classification systems is that tumors may fall under different classes depending on which

International Classification of Retinoblastoma
Group A -tumors <3 mm and away from fovea and optic disc Group B -tumors >3 mm, located at macula/peripapillary region, or with subretinal fluid Group C -tumors with focal vitreous or subretinal seeds within 3 mm of tumor Group D -tumors with diffuse vitreous or subretinal seeds >3 mm away from tumor Group E -tumors covering >50% of globe with or without neovascular glaucoma, hemorrhage, or extension of tumor to optic nerve/anterior chamber system is used, and corresponding treatment choices therefore may not be homogeneous. This is particularly true for group D tumors, whose definitions vary the most. To add another dimension of complexity, treatment choices are also influenced by the financial and medical resources that are available to the ocular oncologist and the patient. IAC has been demonstrated to be an excellent treatment modality with superior tumor control and globe preservation. Numerous studies suggest that IAC can outperform IVC for advanced tumors belonging in groups C and D and help avoid enucleation 21,23 . However, for successful IAC, it often requires a multidisciplinary approach including expert ocular oncologists and neurointerventionalists, who are not ubiquitously available, especially in less-developed countries, from which more than 80% of the total retinoblastoma patients originate 50 . Such disparities lead to limited availability of IAC to only a fraction of the patients. More than half of the physicians in a global survey of retinoblastoma experts by Scelfo et al. reported using IVC as primary treatment for group D tumors, while 27% perform primary enucleation, followed by only 16% who offer IAC as first-line therapy 51 .
In addition to the issue of availability, enucleation and IVC may sometimes make more financial and logistical sense for patients. Enucleation is significantly less costly than IAC and requires less frequent follow-up. Fabian et al. reported that fewer exams under anesthesia (EUA) are required for group D patients whose eyes were enucleated compared to those who underwent IAC 52 . For patients who have difficulty receiving regular check-ups for various reasons, enucleation may serve as a better alternative.
Studies have demonstrated that there exists a significant difference in treatment outcomes for retinoblastoma between advanced and less-developed nations. In a recent cross-sectional study of almost 280 retinoblastoma centers around the globe, retinoblastoma was diagnosed significantly later (30 months vs. 14 months) in less-developed countries, and patients often had advanced disease with extraocular extension of the tumor (49%) as well as metastasis (19%), whereas these were much less frequently observed in advanced nations (1.5% extraocular extension and 0.3% metastasis) at the time of diagnosis 10,53 . Unfortunately, as a result, treatment outcomes are usually significantly worse owing to late presentation in developing countries, and mortality rates are as high as 60-70% in some nations 54-58 . Considering how major advancements in retinoblastoma research have greatly shifted the paradigm of the treatment of retinoblastoma and have made it a largely treatable and curable cancer in advanced nations, there have been efforts to mitigate the mortality and enucleation rates in less-developed countries. Sharing of resources and adequately educating patients and their families about retinoblastoma and its treatment options are some of the ongoing efforts, but there is still certainly a large gap to close 57 .

Genomic analysis of retinoblastoma
As germline retinoblastoma accounts for approximately 30-40% of all retinoblastoma cases, it is important to determine for each patient whether he/she has a mosaic or germline mutation and therefore possesses a risk for secondary malignancies in other parts of the body as well as the possibility of retinoblastoma in other family members. In cases of bilateral disease, a germline mutation can be assumed, but it is inadvisable in cases of unilateral disease to assume a somatic mutation. Studies have shown that approximately 15% of unilateral cases are due to germline mutations, and they can subsequently develop tumors in the unaffected eye 59 . Therefore, early genotyping can provide crucial insights into what type of retinoblastoma the patient has, enabling optimal treatment plans and genetic counseling 60 . Recently, there has been a potentially exciting development of cell-free DNA analysis of  Once this molecular testing method is validated in larger multicenter studies, cell-free DNA analysis of retinoblastoma may allow physicians to precisely identify germline RB1 or MYCN mutations and prepare personalized therapy regimens and family genetic counseling.
New therapeutic targets and future directions For the past two decades, active research has been undertaken to identify new molecular targets specific for retinoblastoma. Some recent publications have identified several genes and pathways that may hold the key for retinoblastoma tumorigenesis and potential therapy options. There is growing evidence that retinoblastoma cells and retinal cone precursor cells may share common pathways for proliferation. Previously, it has been observed that retinoblastoma cells require suppression of p27, which is regulated by increased SKP2 expression 64 , and also that cone precursor cells have downregulated p27 65 . Moreover, it has been observed that, in order to escape apoptosis and further proliferate, retinoblastoma cell proliferation depends on the expression of proteins including RXRγ, MYCN, MDM2, and TRβ2, which are also greatly produced by retinal cone progenitor cells 66 . Xu et al. published that when retinal cone progenitor cells undergo a RB1 gene knock-out, they exhibit significantly increased proliferation and dependence on MYCN, MDM2, and SKP2 expression, which is very similar to how retinoblastoma cells proliferate 67 . Together, it is postulated that retinoblastoma may possibly originate from cone precursors and that targeting some of the key cone precursor cell signaling pathways may allow us to gain insights into specific gene-targeted therapy for retinoblastoma. Other molecular targets such as the spleen tyrosine kinase 68 and orthodenticle homeobox 2 (OTX2) 69 are also being investigated.
With increasing emphasis by researchers on the molecular understanding of retinoblastoma, significant advancements are anticipated in the coming decade.

Uveal melanoma
Uveal melanoma, although rare, is the most common primary intraocular cancer in adults, with an incidence rate of 5-7 cases per million or approximately 2,000 new cases per year in the United States 2 . Uveal melanoma most frequently presents unilaterally in the sixth or seventh decade of life and has a heavy predominance in the Caucasian population 2 . Very rarely, patients with germline BAP1 mutations can develop bilateral primary uveal melanoma 70 . Current mainstays of treatment are plaque brachytherapy (most commonly iodine-125 in the United States and ruthenium-106 in Europe and other nations) or proton beam irradiation (PBI). Advanced cases may require enucleation, while some localized iris tumors, which account for only 3-5% of all cases, may be surgically excised 71 . While the overall 5-year survival rate has remained close to 80% 2,72 , up to 60% of uveal melanoma patients develop metastatic disease despite timely treatment and consistent follow-up 73,74 . This is thought to be due to early hematogenous spreading of the tumor cells that are often difficult to detect with current diagnostic capacities 75 . Liver, lung, and bone are the most common organs to which uveal melanoma metastasizes, and once metastatic disease is identified, there is not yet an effective treatment that can meaningfully lengthen survival in all patients 73,74,[76][77][78][79] .

Genomic analyses of uveal melanoma
Numerous studies have helped us better understand how uveal melanoma develops and how it differs from a choroidal nevus, which is a benign intraocular lesion that is commonly seen in up to 10% of the population, especially among Caucasians 80 . Most nevi are found in the posterior half of the eye and remain stable in size or grow very slowly over many years 81 . On the other hand, choroidal melanoma appears to undergo several critical genomic and molecular changes that are quite different from the mutations seen in cutaneous melanoma, such as the BRAF mutation 82 . The pathogenesis of uveal melanoma usually involves an initiating mutation in the Gα11/Q pathway, which activates multiple downstream pathways including the MAPK and YAP 83 pathways. Choroidal nevi have GNAQ or GNA11 mutations too, but uveal melanoma further undergoes one of three key driver mutations as part of malignant transformation: BAP1, SF3B1, or EIF1AX. BAP1 is a tumor suppressor gene located on chromosome 3 that plays a role in protein de-ubiquitination, cell cycle regulation, and DNA repair 84,85 . Specifically, BAP1 interacts with genes that are critical for maintaining the differentiated state of the cell, and a BAP1 mutation can thus make the cell become more stem cell-like and ultimately increase the risk of metastasis in uveal melanoma 86 . Either monosomy 3 or a somatic knock-out of the second BAP1 allele leads to BAP1 inactivation in approximately 35-45% of all uveal melanomas and confers the worst prognosis of these three key driver genes 87,88 . SF3B1 is a splicing gene whose mutation accounts for approximately 20-25% of uveal melanoma cases 87,89 . When a mutation occurs in the SF3B1 gene, which normally codes for the U2 snRNP complex of the spliceosome, frameshift insertions may occur and lead to incorrect splicing of the pre-mRNA and subsequent mRNA degradation 89,90 . It is, however, not yet fully understood how this spliceosome aberration is linked to the malignant transformation of uveal melanoma. Lastly, EIF1AX is a gene that encodes the eIF1A initiation factor, which is a component of the 43S preinitiation complex that is involved in the early steps of translation initiation. Mutations in EIF1AX, seen in approximately 20-25% of tumors, lead to amino acid substitutions or deletions in the N-terminal tail and interfere with proper translation 88,91 . In most cases, mutations of these three key genes occur in a mutually exclusive fashion, and the specific prognostication differs based on which mutation the tumor harbors.
Multiple studies have suggested that BAP1, EIF1AX, and SF3B1 mutations are linked to other prognostic indicators. Gene expression profiling (GEP), which quantifies the mRNA expression of 12 key genes (and three control genes), classifies uveal melanoma as either class 2 with high metastatic risk or class 1 with lower metastatic risk 92,93 . Class 1 tumors are further subdivided into class 1A, which carries a 2% risk of metastases at 5 years, and class 1B, which carries a 21% risk of metastases at 5 years 94 . Although class 1B tumors do not metastasize as aggressively as class 2 tumors, which have a 72% incidence of metastatic disease at 5 years, 1B tumors are thought to carry a gradually increasing metastatic risk over time 94 . Therefore, close monitoring of class 1B and class 2 tumors with abdominal imaging is recommended 94 . Current data suggest that BAP1 is predominantly seen in tumors with a GEP class 2 signature, loss of chromosome 3, and gain of chromosome 8q and leads to significantly worse prognosis 95 . SF3B1 is more closely associated with class 1B tumors and leads to late-onset metastasis, conferring intermediate risk. EIF1AX mutations are generally seen in patients with a class 1A signature and are associated with a lower metastatic risk 94,96 . Therefore, once uveal melanoma is diagnosed, utilization of prognostic molecular testing can help guide the patient to receive personalized disease monitoring.
Although our understanding of primary uveal melanoma has greatly improved over the past two decades owing to major advancements in genomic and molecular research, there is still much to learn about metastatic uveal melanoma because no new therapies have decisively reduced the mortality rate. Nonetheless, recently, there have been several new and potentially controversial reports on the genomics of the disease. Shain et al. reported that some metastatic lesions could potentially develop even before key chromosomal or genomic changes occur in the primary tumor, including the 8q gain or loss of both BAP1 alleles 97 . In the metastatic lesions, new oncogenic changes, such as CDKN2A loss, MBD4 deficiency, and EZH2, PTK2B, and PBRM1 gene mutations, were observed 97-99 , indicating that primary and metastatic uveal melanoma may have quite different genetic compositions. This data has not been validated on a larger scale in additional laboratories.

Plaque brachytherapy and proton beam irradiation
The Collaborative Ocular Melanoma Study (COMS) published a series of landmark studies on uveal melanoma since 1990 that laid the cornerstone for diagnosis and management of uveal melanoma. The pivotal COMS medium tumor trial compared the efficacy of iodine-125 plaque brachytherapy with that of enucleation and concluded that there was no statistically significant difference in survival between patients who underwent plaque brachytherapy and patients who underwent enucleation 100 . Since then, most centers have adopted plaque brachytherapy as the standard treatment for uveal melanoma, and iodine-125 brachytherapy is the most commonly used treatment modality for uveal melanoma in the United States with excellent clinical outcomes (Figure 2). Outside the United States, ruthenium-106 and palladium-103 are also frequently used 101 . Iodine-125 and palladium-103 are both gamma ray emitters with a lower energy profile, while ruthenium-106 emits beta rays and has a steeper dose fall-off 102 . With the steeper dose fall-off, ruthenium-106 may cause less radiation-induced damage to the surrounding normal structures that are important for the preservation of vision, including the macula and optic disc, whereas iodine-125 can be effective for thicker tumors (>5 mm in height) for which ruthenium-106 may not be adequate to achieve local control 103 .
PBI has also been used as primary therapy for uveal melanoma for several decades. Tantalum beads (MRI-safe metal) are surgically attached to the sclera, and the beads can remain in the orbit permanently, even after PBI. A series of PBI sessions occur over several days with a total cumulative radiation dosage of 50 to 70 Gray (Gy) 104 . In addition to small and medium-sized tumors, PBI can be particularly effective for patients with large tumors that sometimes may not be ideal candidates for plaque brachytherapy owing to size limitations, albeit typically with severe long-term radiation complications. Papakostas and colleagues published a single center cohort study of patients with large tumors (mean largest basal diameter [LBD] of 18.1 mm and thickness of 8.2 mm) who were treated with PBI; 10-year eye retention rate was 70.4%, and tumor control rate was 87.5% 105 .
At advanced centers, PBI and plaque brachytherapy have comparable tumor control rates, with multiple recent publications reporting local recurrence rates of 5% or less for both treatment modalities [106][107][108][109][110][111][112] . Some reports even demonstrated no local treatment failure for up to 2 years post-brachytherapy 110 , which is a significant improvement from the 10.3% local recurrence rate at 5 years after brachytherapy that was originally reported in the COMS publications 113 . Achieving long-term local tumor control is critical because the probability of the primary lesion metastasizing significantly increases once tumor recurrences are observed 111,114 . One study by Bellerive et al. reported a local recurrence rate of 5.6% with a median onset of 18 months after initial brachytherapy 108 . In this study, one-third of the patients with local recurrences underwent repeat brachytherapy with no further evidence of recurrence afterwards during a median follow-up of 45 months. In another study by Riechardt et al., salvage PBI was performed for 48 patients who developed local recurrences after the first therapy, with 92.1% local control at 10 years of follow-up 115 . However, even though local control was achieved in most cases, both studies had a 30-50% rate of metastatic disease over 10 years, demonstrating again that the risk of metastatic mortality is significantly elevated when local treatment failure occurs.
Several key factors have contributed to improved treatment outcomes for both PBI and plaque brachytherapy. Detailed imaging studies including fundus photographs, computed tomography (CT), and ultrasonography allow for accurate determination of the tumor anatomy, tumor location, as well as pre-treatment 3D planning 116 . Based on this information, radiation dosage simulation is performed, and potential radiation exposure to the critical structures including the optic nerve and macula is also calculated 104,110 . In addition to the 3D simulation, intraoperative ultrasound confirmation of the plaque placement, which has been in use for the past two decades 117-119 , serves a critical role to ensure that all tumor margins are fully covered by the plaque that is newly inserted. Aziz et al. published a retrospective study in which the patients who underwent intraoperative ultrasonographic confirmation of the plaque position experienced a 1.5% treatment failure in contrast to the 9.3% in those who did not 120 . For plaque brachytherapy, thinner and customizable designs also help conform to each tumor's specific geometry and offer more targeted therapy while minimizing radiation damage to critical ocular structures 121-123 .

Radiation retinopathy
Patients undergoing PBI or plaque brachytherapy are at risk of developing radiation-related ocular complications based on which ocular structures the radiation beam traverses. Anterior segment complications include cataract, rubeosis iridis, and neovascular glaucoma 124 . Posteriorly, optic neuropathy and radiation retinopathy are common. Radiation damages capillary endothelial cells and pericytes, which then causes areas of retinal capillary nonperfusion 125 . Severe retinal ischemia can then subsequently lead to neovascularization and retinal hemorrhage as well as microaneurysms that can affect visual acuity. Tumors that are within 4 mm of the fovea or larger than 10 mm in LBD are at significant risk for radiation maculopathy 126 . As high as 50% of patients undergoing radiotherapy for uveal melanoma may develop OCT-or OCT-A evident macular edema, which is an early sign of radiation retinopathy that typically manifests between 1 and 2 years post-radiotherapy [127][128][129] .
Treatment options for radiation retinopathy include laser photocoagulation, intravitreal anti-VEGF, and steroids. Most studies investigating the use of intraocular steroids have been limited by small sample sizes and lack of randomization [130][131][132][133] .
On the other hand, multiple studies have investigated the use of intravitreal anti-VEGF and demonstrated efficacy for preserving best-corrected visual acuity (BCVA). In addition to numerous retrospective studies reporting on improved macular edema and central macular thickness (CMT) after the use of anti-VEGF therapy, several key prospective studies have shown statistically significant improvement in BCVA after regularly injecting intravitreal anti-VEGF. Murray et al. reported that 39 patients who received aflibercept injections either at a fixed 6-week interval or via a treat-and-extend protocol after developing post-radiotherapy radiation maculopathy showed improvements in BCVA at 1 year follow-up, with 42.5% having better than 20/50 vision, as well as enhanced CMT 134 . Fallico, Russo, and colleagues, despite a small sample size of nine eyes, also reported similar improvements in BCVA and CMT after administering a mean of 4.4 aflibercept injections over 2 years 135 . Schefler et al. published results of a prospective, multicenter ranibizumab therapy trial for clinically evident macular edema after iodine-125 plaque brachytherapy 136 . The study compared a monthly ranibizumab injection group to two other groups: one who received targeted pan-retinal laser photocoagulation (TRP) with monthly ranibizumab and another group that received as-needed Similarly, Seibel et al. reported that patients who received four to six ranibizumab injections in the first 6 months after developing radiation maculopathy had significantly better BCVA than a control group that received pan-retinal laser photocoagulation to the ischemic retina 138 . Kim et al. 139 published a prospective study in which 40 patients received prophylactic ranibizumab every 2 months for 2 years starting 2 weeks prior to PBI therapy for uveal melanoma. In this study, 88% of patients maintained 20/40 or better visual acuity at 2 years of follow-up, despite the fact that all of the study patients had tumors within two disc diameters of the macula or the optic nerve and therefore had high risk for developing radiation-induced maculopathy or papillopathy.
The exact timing of anti-VEGF therapy initiation for radiation retinopathy is not yet known, as both prophylactic injections 139,140 and injections once clinically evident macular edema manifests 128,136 have demonstrated clinical efficacy. Nonetheless, the current general consensus is that, in order to effectively maintain or improve visual acuity and reduce macular edema, long-term anti-VEGF injections need to be administered on a regular basis.
Clinical and molecular features with prognostic significance Molecular analysis methods including the gene expression profile (GEP) and multiplex ligation-dependent probe amplification (MLPA) can give valuable prognostic information by identifying key genomic mutations and corresponding metastatic potential. Various clinical risk factors, when combined with the genomic classifications, may strengthen the prognostic predictions for metastatic disease in uveal melanoma. Tumor size has long been understood to be a key factor for predicting outcomes. The LBD can serve as an independent marker for predicting metastasis 141,142  reported that patients with class 2 uveal melanoma had a significantly lower 5-year metastasis-free survival when the tumor was larger than 12 mm in LBD, which was the same LBD threshold for a worse prognosis in a study by Demirci et al. 142 .
Preferentially expressed antigen in melanoma (PRAME) is a cancer-testis antigen that is not normally produced by healthy human tissues but is expressed in different types of cancer 146 . PRAME helps cancer cells to survive by suppressing retinoic acid-driven apoptosis and inhibition of cell proliferation 147 . Field et al. reported that increased PRAME mRNA expression is observed in uveal melanoma and that GEP class 2 tumors are more likely than the class 1 signature to be PRAME+ 148 . Furthermore, PRAME+ class 2 tumors progress faster to metastases, and PRAME+ class 1 tumors are more likely to metastasize than PRAME-tumors 149 . Newer publications have investigated possible relationships between the PRAME status and other prognostic factors, and more studies on this subject are in progress. Cai et al. reported that the combined PRAME status and GEP information can offer a higher prognostic accuracy than optimized TNM staging alone 150 . Schefler et al. conducted a retrospective study in which PRAME status was analyzed in accordance with other clinical and molecular factors 151 . In 148 uveal melanoma patients who underwent PRAME testing, 37% of the patients were positive for PRAME and subsequently had a statistically significant association with a worse GEP classification. More than half of GEP class 2 tumors were PRAME+, while approximately 30% of the class 1 tumors were PRAME+. Given that certain class 1A tumors still progress to metastatic disease and that PRAME+ appears to be associated with increased metastatic disease rates, a positive PRAME expression status in a GEP class 1 tumor may explain the rare cases in this group that do metastasize. Future studies will help identify additional associations between PRAME and other more established risk factors for metastasis.
There has been a debate on whether tumor response to radiotherapy can be predicted based on the GEP classification. While several clinicians have reported no significant difference in the rate of regression between class 1 and class 2 tumors at 3 and 6 months after iodine-125 brachytherapy 152,153 , Rao et al. 154 and Mruthyunjaya et al. 155 observed significantly greater reductions in tumor thickness in class 1 tumors at 3 and 6 months post-therapy and 3 months post-therapy, respectively. More data are needed on this subject to elucidate a clearer relationship.
Recently, Liu et al. published an interesting pilot study introducing deep machine learning to correlating histopathologic significance to the GEP classification 156 . In this study, machine learning assessed 10 hematoxylin-eosin pathology slides from GEP class 1 tumors and 10 from class 2 tumors based on classic tumor features: spindle-type cells for lower-grade tumors and epithelioid-type cells with high degree of atypia for higher-grade tumors. Interestingly, 15 of the 20 slides were correctly analyzed and classified as either GEP class 1 or 2.
Although the data from this study are limited and preliminary, this result suggests that deep machine learning can potentially be applied in various ways to assist clinicians with predicting metastatic risk in uveal melanoma.

Other therapies
Because our understanding of the molecular genetics of uveal melanoma has seen rapid advancements in the past decade, clinical trials utilizing our understanding of these pathways have begun. Durante and colleagues recently reported that tumor-infiltrating T cells in uveal melanoma express not just PD1 and CTLA4 checkpoint markers but also LAG3, which plays a critical role in T cell exhaustion and subsequent immune escape 157 . In their study, class 2 tumors predominantly expressed LAG3, not PD1 or CTLA4, which may account for why checkpoint inhibitor therapy targeting PD1 and CTLA4 fails to work in many cases. Faião-Flores et al. investigated various cell signaling pathways in uveal melanoma and observed YAP and AKT upregulations when the MEK pathway was inhibited, explaining the increased resistance by uveal melanoma cells upon MEK inhibition 158 . Interestingly, when panobinostat, a histone deacetylase (HDAC) inhibitor, was added to the MEK inhibitors, the drug appeared to limit the MEK resistance by downregulating the YAP and AKT signaling. If further studies and clinical trials show promising outcomes, immunotherapy that targets the LAG3 checkpoint or the HDAC/MEK pathway may provide new insights into uveal melanoma treatment. Such trials are now being planned.
Recently, Middleton et al. 159 published a phase I/II multicenter clinical trial of a bispecific fusion protein named tebentafusp that targets gp100, a melanocytic antigen that is expressed in both cutaneous and uveal melanoma. Tebentafusp can bind the gp100 protein presented on HLA-A*02 via its T cell receptor domain with a high affinity, while its anti-CD3 antibody domain binds surveilling CD3 + T cells to generate an immune response against the melanoma cells. The study's metastatic uveal melanoma patients (19 out of 84 study patients) achieved a 1-year survival rate of 65% 159 . In vitro studies suggest that as few as 10 epitopes are needed to generate a sufficient immune response against melanoma cells, demonstrating tebentafusp's high specificity for gp100 and its potential role in metastatic uveal melanoma 160,161 . A phase III clinical trial of tebentafusp will be completed in the near future.
In addition to immunotherapy, another therapeutic modality that is currently undergoing a phase Ib clinical trial is a nanoparticle therapy that uses photosensitive nanoparticles that preferentially bind tumor cells 162 . Once light-activated, these particles can selectively kill tumor cells with minimal damage to the surrounding normal tissues. This treatment modality, if proven successful in clinical trials, has the potential to preserve much of the patient's vision and could be particularly groundbreaking in patients with small tumors that are close to critical ocular structures such as the optic nerve and the macula.

Conclusion
The field of ocular oncology has seen major advancements in both diagnostic and treatment modalities for retinoblastoma and uveal melanoma in the last two decades. For retinoblastoma, the paradigm has shifted toward minimally invasive, targeted interventions, namely IAC and IVitC, which have dramatically turned retinoblastoma into a curable and manageable cancer in developed countries. On the other hand, innovations in the genomic analysis of uveal melanoma have allowed ocular oncologists to better characterize each tumor based on specific mutations and to predict the disease course and treatment outcomes. Although metastatic uveal melanoma remains a critical challenge for which no successful cure has been introduced, our conjoined efforts to detect and treat uveal melanoma at an early stage and to improve the treatment outcome of metastatic disease will hopefully come to fruition in the near future.