Predicting Choroidal Nevus Transformation to Melanoma Using Machine Learning.

Purpose: To develop and validate machine learning (ML) models to predict choroidal nevus transformation to melanoma based on multimodal imaging at initial presentation. Design: Retrospective multicenter study. Participants: Patients diagnosed with choroidal nevus on the Ocular Oncology Service at Wills Eye Hospital (2007-2017) or Mayo Clinic Rochester (2015-2023). Methods: Multimodal imaging was obtained, including fundus photography, fundus autofluorescence, spectral domain OCT, and B-scan ultrasonography. Machine learning models were created (XGBoost, LGBM, Random Forest, Extra Tree) and optimized for area under receiver operating characteristic curve (AUROC). The Wills Eye Hospital cohort was used for training and testing (80% training-20% testing) with fivefold cross validation. The Mayo Clinic cohort provided external validation. Model performance was characterized by AUROC and area under precision-recall curve (AUPRC). Models were interrogated using SHapley Additive exPlanations (SHAP) to identify the features most predictive of conversion from nevus to melanoma. Differences in AUROC and AUPRC between models were tested using 10 000 bootstrap samples with replacement and results. Main Outcome Measures: Area under receiver operating curve and AUPRC for each ML model. Results: There were 2870 nevi included in the study, with conversion to melanoma confirmed in 128 cases. Simple AI Nevus Transformation System (SAINTS; XGBoost) was the top-performing model in the test cohort [pooled AUROC 0.864 (95% confidence interval (CI): 0.864-0.865), pooled AUPRC 0.244 (95% CI: 0.243-0.246)] and in the external validation cohort [pooled AUROC 0.931 (95% CI: 0.930-0.931), pooled AUPRC 0.533 (95% CI: 0.531-0.535)]. Other models also had good discriminative performance: LGBM (test set pooled AUROC 0.831, validation set pooled AUROC 0.815), Random Forest (test set pooled AUROC 0.812, validation set pooled AUROC 0.866), and Extra Tree (test set pooled AUROC 0.826, validation set pooled AUROC 0.915). A model including only nevi with at least 5 years of follow-up demonstrated the best performance in AUPRC (test: pooled 0.592 (95% CI: 0.590-0.594); validation: pooled 0.656 [95% CI: 0.655-0.657]). The top 5 features in SAINTS by SHAP values were: tumor thickness, largest tumor basal diameter, tumor shape, distance to optic nerve, and subretinal fluid extent. Conclusions: We demonstrate accuracy and generalizability of a ML model for predicting choroidal nevus transformation to melanoma based on multimodal imaging. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Thirteen-Year Follow-Up of Conjunctival Amyloidosis.

This case report discusses a diagnosis of conjunctival amyloidosis in an otherwise asymptomatic male aged 71 years with 13-year follow-up.

Impact of Tumor Pigmentation in 6934 Patients with Uveal Melanoma at a Single Center.

Purpose: To evaluate clinical features and outcomes associated with degree of tumor pigmentation in patients with uveal melanoma (UM) of the choroid and ciliary body. Design: Retrospective observational study. Subjects: Six thousand nine hundred thirty-four consecutive patients with choroidal or ciliary body melanoma between 1971 and 2007 from a single ocular oncology center. Methods: Data on patient demographics, tumor characteristics, treatment approach, and clinical outcomes were collected. Comparisons between pigmented (>80% pigmentation by surface area), partially pigmented (20%-80%), and nonpigmented tumors (<20%) were performed using relevant hypothesis testing. Survival analyses for metastasis and melanoma-related death were conducted using the Kaplan-Meier method with log-rank tests for univariate comparisons. A multivariate Cox regression analysis was performed to assess the independent effects of multiple covariates on time-to-metastasis. Main Outcome Measures: Extraocular extension, ocular melanocytosis, time to tumor recurrence, tumor location, and melanoma-related metastasis and death. Results: There were 6934 eyes with UM and the degree of tumor pigmentation was classified as pigmented (n = 3762; 54%), partially pigmented (n = 2115; 31%), or nonpigmented (n = 1057; 15%). Pigmented UM was associated with extraocular extension (P < 0.001), ocular melanocytosis (P = 0.003), earlier tumor recurrence (P < 0.001), and more anterior tumor epicenter location (ciliary body, and equator to ora serrata) (P < 0.001). Pigmented UMs also exhibited the highest 10-year metastasis rate at 26%, compared with 19% for partially pigmented UMs and 16% for nonpigmented UMs (P < 0.001). Kaplan-Meier survival curves demonstrated differences among the tumor pigmentation groups for melanoma-related metastasis (P < 0.001) and melanoma-related death (P < 0.001). Multivariate Cox regression analysis for melanoma-related metastasis showed that pigmented UMs had a 29% higher relative risk of developing metastasis compared with partially pigmented UMs (P = 0.002) and a 54% higher relative risk of developing metastasis compared with nonpigmented UMs (P < 0.001). Conclusions: Pigmented choroidal and ciliary body melanoma is more often associated with ocular melanocytosis, extraocular extension, anterior tumor epicenter, and earlier tumor recurrence. We also revealed that patients with pigmented UMs demonstrate a higher 10-year rate of metastatic disease and have decreased metastatic survival relative to partially pigmented and nonpigmented UMs. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Retinoblastoma Outcomes Based on the 8th Edition American Joint Committee on Cancer Pathological Classification in 1411 Patients.

PURPOSE: To evaluate the outcomes of retinoblastoma (RB) based on the 8th edition of the American Joint Committee on Cancer (AJCC) pathological classification in a global cohort of patients. DESIGN: Retrospective, multicentre, intercontinental collaborative study PARTICIPANTS: 1411 patients INTERVENTION(S): Primary enucleation with/without adjuvant chemotherapy/radiotherapy MAIN OUTCOMES(S): Orbital tumor recurrence, tumor-related metastasis, tumor-related death RESULTS: Based on the 8th edition AJCC pathological classification, 645 (46%) eyes belonged to pT1, 164 (11%) to pT2, 493 (35%) to pT3, and 109 (8%) to pT4 categories. At a mean follow-up of 38 months (median, 35 months; <1-149 months), orbital tumor recurrence was seen in 8 (1%), 5 (3%), 22 (4%) and 25 (23%) of pT1, pT2, pT3, and pT4 (p<0.001) categories, respectively; tumor-related metastasis was seen in 7 (1%), 5 (3%), 40 (8%), and 46 (43%) of pT1, pT2, pT3, and pT4 (p<0.001) categories, respectively; tumor-related death was seen in 12 (2%), 7 (4%), 64 (13%), and 64 (59%) of pT1, pT2, pT3, and pT4 (p<0.001) categories, respectively. Multivariate Cox proportional hazards analysis of outcomes revealed pT category and adjuvant therapy as independent predictors of outcomes. Categories pT3b (p=0.005), pT3c (p<0.001), pT3d (p<0.001), and pT4 (p<0.001) had a greater hazard for orbital recurrence; categories pT2a (p=0.015), pT3a (p<0.001), pT3b (p<0.001), pT3c (p<0.001), pT3d (p<0.001) and pT4 (p<0.001) had a greater hazard for tumor-related metastasis; and categories pT2a (p=0.068), pT2b (p=0.004), pT3a (p<0.001), pT3b (p<0.001), pT3c (p<0.001), pT3d (p<0.001) and pT4 (p<0.001) had a greater hazard for tumor-related death when compared to the pT1 category. Patients who did not receive adjuvant therapy had greater hazards of orbital tumor recurrence in categories pT3b (p=0.005), pT3c (p=0.003), and pT4 (p=0.002); greater hazards of tumor-related metastasis in categories pT3a (p=0.001), pT3b (p=0.01), pT3c (p=0.001), and pT4 (p=0.007); and tumor-related death in categories pT3a (p<0.001), pT3b (p=0.009), pT3c (p=0.018), and pT4 (p<0.001) when compared to those who received adjuvant therapy. CONCLUSION: The 8th edition AJCC pathological classification predicts outcomes in patients undergoing primary enucleation for RB, and adjuvant therapy is associated with a lower risk of orbital recurrence, tumor-related metastasis, and tumor-related death in the pT3 and pT4 categories.

Genome-Wide Methylation Patterns in Primary Uveal Melanoma: Development of MethylSig-UM, an Epigenomic Prognostic Signature to Improve Patient Stratification.

Despite studies highlighting the prognostic utility of DNA methylation in primary uveal melanoma (pUM), it has not been translated into a clinically useful tool. We sought to define a methylation signature to identify newly diagnosed individuals at high risk for developing metastasis. Methylation profiling was performed on 41 patients with pUM with stage T2-T4 and at least three years of follow-up using the Illumina Infinium HumanMethylation450K BeadChip (N = 24) and the EPIC BeadChip (N = 17). Findings were validated in the TCGA cohort with known metastatic outcome (N = 69). Differentially methylated probes were identified in patients who developed metastasis. Unsupervised consensus clustering revealed three epigenomic subtypes associated with metastasis. To identify a prognostic signature, recursive feature elimination and random forest models were utilized within repeated cross-validation iterations. The 250 most commonly selected probes comprised the final signature, named MethylSig-UM. MethylSig-UM could distinguish individuals with pUM at diagnosis who develop future metastasis with an area under the curve of ~81% in the independent validation cohort, and remained significant in Cox proportional hazard models when combined with clinical features and established genomic biomarkers. Altered expression of immune-modulating genes were detected in MethylSig-UM positive tumors, providing clues for pUM resistance to immunotherapy. The MethylSig-UM model is available to enable additional validation in larger cohort sizes including T1 tumors.

High-Risk Histopathological Features of Retinoblastoma following Primary Enucleation: A Global Study of 1426 Patients from 5 Continents.

PURPOSE: To evaluate high-risk histopathological features (HRHF) following primary enucleation of eyes with retinoblastoma (RB) and assess the patient outcomes across continents. METHODS: Retrospective study of 1426 primarily enucleated RB eyes from five continents. RESULTS: Of all, 923 (65%) were from Asia (AS), 27 (2%) from Australia (AUS), 120 (8%) from Europe (EUR), 162 (11%) from North America (NA), and 194 (14%) from South America (SA). Based on the continent (AS vs. AUS vs. EUR vs. NA vs. SA), the histopathology features included massive choroidal invasion (31% vs. 7% vs. 13% vs. 19% vs. 27%, p=0.001), post-laminar optic nerve invasion (27% vs. 0% vs. 16% vs. 21% vs. 19%, p=0.0006), scleral infiltration (5% vs. 0% vs. 4% vs. 2% vs. 7%, p=0.13), and microscopic extrascleral infiltration (4% vs. 0% vs. <1% vs. <1% vs. 4%, p=0.68). Adjuvant chemotherapy with/without orbital radiotherapy was given in 761 (53%) patients. Based on Kaplan-Meier estimates in different continents (AS vs. AUS vs. EUR vs. NA vs. SA), the 6-year risk of orbital tumor recurrence was 5% vs. 2% vs. 0% vs. 0% vs. 12% (p<0.001), systemic metastasis was reported in 8% vs. 5% vs. 2% vs. 0% vs. 13% (p=0.001), and death in 10% vs. 3% vs. 2% vs. 0% vs. 11% (p<0.001) patients. CONCLUSION: There is a wide variation in the infiltrative histopathology features of RB across continents, resulting in variable outcomes. SA and AS had a higher risk of orbital tumor recurrence, systemic metastasis, and death compared to AUS, EUR, and NA.

Targeting GPC2 on Intraocular and CNS Metastatic Retinoblastomas with Local and Systemic Delivery of CAR T Cells.

PURPOSE: Retinoblastoma is the most common intraocular malignancy in children. Although new chemotherapeutic approaches have improved ocular salvage rates, novel therapies are required for patients with refractory intraocular and metastatic disease. Chimeric antigen receptor (CAR) T cells targeting glypican-2 (GPC2) are a potential new therapeutic strategy. EXPERIMENTAL DESIGN: GPC2 expression and its regulation by the E2F1 transcription factor were studied in retinoblastoma patient samples and cellular models. In vitro, we performed functional studies comparing GPC2 CAR T cells with different costimulatory domains (4-1BB and CD28). In vivo, the efficacy of local and systemic administration of GPC2 CAR T cells was evaluated in intraocular and leptomeningeal human retinoblastoma xenograft models. RESULTS: Retinoblastoma tumors, but not healthy retinal tissues, expressed cell surface GPC2, and this tumor-specific expression was driven by E2F1. GPC2-directed CARs with 4-1BB costimulation (GPC2.BBz) were superior to CARs with CD28 stimulatory domains (GPC2.28z), efficiently inducing retinoblastoma cell cytotoxicity and enhancing T-cell proliferation and polyfunctionality. In vivo, GPC2.BBz CARs had enhanced persistence, which led to significant tumor regression compared with either control CD19 or GPC2.28z CARs. In intraocular models, GPC2.BBz CAR T cells efficiently trafficked to tumor-bearing eyes after intravitreal or systemic infusions, significantly prolonging ocular survival. In central nervous system (CNS) retinoblastoma models, intraventricular or systemically administered GPC2.BBz CAR T cells were activated in retinoblastoma-involved CNS tissues, resulting in robust tumor regression with substantially extended overall mouse survival. CONCLUSIONS: GPC2-directed CAR T cells are effective against intraocular and CNS metastatic retinoblastomas.

Vitreoretinal Lymphoma Masquerading as Central Serous Chorioretinopathy.

The following is a case of vitreoretinal lymphoma masquerading as central serous chorioretinopathy (CSCR). A 74-year-old man presented with blurred vision in the left eye with unilateral subretinal fluid in the setting of exogenous corticosteroid use, which was diagnosed as CSCR and resolved with corticosteroid cessation. He later experienced a similar self-limited episode in the right eye. Subsequently, he developed bilateral vitritis with yellow-white subretinal pigment epithelial infiltrates. Vitreous biopsy confirmed a diagnosis of large B-cell lymphoma. Vitreoretinal lymphoma can masquerade as a number of ocular pathologies, including CSCR. [Ophthalmic Surg Lasers Imaging Retina 2024;55:467-470.].

Multifocal Unilateral Orange Fundus Tumors in a Young Man.

A 38-year-old man had an asymptomatic, orange, macular choroidal mass with macular choroidal folds and a retinal pigment epithelial detachment in the right eye and a second orange mass nasal to the optic disc also in the right eye. What would you do next?

Topical 5-fluorouracil 1% for moderate to extensive ocular surface squamous neoplasia in 73 consecutive patients: Primary versus secondary treatment.

IMPORTANCE: Ocular surface squamous neoplasia (OSSN) is a spectrum of malignancies that generally includes conjunctival intraepithelial neoplasia (CIN) and squamous cell carcinoma (SCC). OSSN can be treated with topical therapies including interferon α-2b (IFN), mitomycin C (MMC), or 5-fluorouracil 1% (5FU). Recently, due to unavailability of IFN and toxicity associated with MMC, therapy has shifted towards 5FU. OBJECTIVE: Herein, we compare the use of 5FU 1% as a primary versus (vs) secondary treatment regimen in eyes with moderate to extensive OSSN. DESIGN SETTING AND PARTICIPANTS: Retrospective cohort study of 73 consecutive patients with unilateral moderate to extensive OSSN treated at a single tertiary ocular oncology center from 2016 to 2023. Mean follow up time was 478.2 days overall, with 283.0 days for primary 5FU group and 860.3 days for secondary 5FU group. INTERVENTION: Topical 5FU 1% 4 times daily for 2 weeks with option for 2-weekly extension until tumor control, either as primary treatment or as secondary treatment to surgical resection, topical IFN or topical MMC, or cryotherapy. MAIN OUTCOMES: Outcome measures included tumor response, need for additional surgery, complications, and visual outcomes. RESULTS: A comparison (primary vs secondary treatment) revealed no difference in mean tumor basal dimension (19.6 vs 17.2 mm, P = 0.46), thickness (3.7 vs 3.4 mm, P = 0.64), or tumor extent (4.4 vs 4.5 clock hours, P = 0.92). The primary treatment group showed greater complete tumor control (77% vs 38%, P = 0.04). Multivariable analysis comparison (primary vs secondary treatment) showed primary treatment more likely to achieve complete tumor control (P = 0.01). There was no difference in the complication rate from 5FU treatment between the groups. There was no difference in visual outcome, and no tumor-related metastasis (0%) or death (0%). CONCLUSION AND RELEVANCE: Topical 5FU 1% is efficacious and safe as a primary or secondary treatment for moderate to extensive OSSN. Tumors treated with primary 5FU 1% demonstrated more complete resolution. In patients with moderate to extensive OSSN, primary treatment with topical 5FU 1% may be warranted.