Modeling Survival Outcomes in Eye Cancer Patients Using Advanced Statistical and Machine Learning Approaches

Abstract

Although uncommon, eye cancers present a serious threat to both vision and overall survival. These malignancies include retinoblastoma, uveal melanoma, and primary intraocular lymphoma. This study employs advanced survival-analysis techniques to explore prognostic factors and to model patient outcomes using data from 5,000 clinical cases. The Kaplan–Meier estimator, Cox proportional-hazards regression, and Random Survival Forest algorithms were applied to estimate survival probabilities and compare model performance. Key determinants of survival included age, tumor stage, gender, and specific genetic markers. Machine-learning models demonstrated superior predictive accuracy relative to traditional methods, suggesting significant potential for individualized prognosis and therapy planning. The findings highlight the value of combining classical statistics with ensemble learning to enhance ocular-oncology analytics and improve patient-specific care strategies.

Introduction

Ocular cancers—such as retinoblastoma, uveal melanoma, and intraocular lymphoma—are rare but critical due to their impact on vision and life expectancy. Prognosis varies based on age, tumor characteristics, and molecular profiles. While recent advancements in machine learning (ML) have improved survival prediction, many models still fail to capture nonlinear relationships among prognostic variables.

???? Study Objectives

1. Analyze survival trends in eye cancer patients.

2. Identify key prognostic factors.

3. Compare traditional statistical methods with machine learning models.

4. Develop an integrated, personalized survival prediction model.

? Problem Statement

Current prognostic tools in ocular oncology are often inaccurate and limited in their ability to capture complex interactions within patient data. There's a need for data-driven models that can accommodate censored data, population diversity, and higher-order variable interactions.

???? Methodology

A. Data

• Retrospective cohort: 5,000 patients

• Included demographics, tumor details, and genetic markers

• Missing data handled via multivariate imputation

B. Analytical Models

1. Kaplan–Meier Estimation

   • Estimated survival probabilities by age, cancer type, and stage

2. Cox Proportional-Hazards Model

   • Identified statistically significant predictors (e.g., age, tumor stage, genetics)

   • Assumptions validated using Schoenfeld residuals

3. Random Survival Forest (RSF)

   • Ensemble ML method using 1,000 trees

   • Evaluated using C-index and Brier score

   • Captured complex, nonlinear interactions in data

C. Model Evaluation

• C-index (discrimination ability)

• Brier score (prediction accuracy)

• Log-rank tests for survival curve comparison

???? Results

• Kaplan–Meier curves showed variation across cancer types:

  • Retinoblastoma patients had longest survival.

• Cox Model Findings:

  • Age, tumor stage, and genetic features significantly influenced survival (p < 0.05).

• RSF Performance:

  • Achieved C-index = 0.896 (higher than Cox model ≈ 0.85)

  • Identified top predictors:

    • Age, tumor stage, and specific genetic variants (some protective)

Conclusion

This study confirms that advanced survival-analysis techniques can substantially improve prognostic modeling in ocular oncology. The Random Survival Forest model produced the most accurate survival estimates, supporting its application in personalized treatment planning. Future research should aim to integrate molecular-omics data and validate predictive models across diverse demographic populations to enhance generalizability and clinical adoption.

References

[1] Bilbeisi, S. C., Brichko, T., Chakraborty, R., et al. (2023). Causes of death and survival analysis for patients with retinoblastoma. Frontiers in Medicine, 10, 1244308. [2] Cox, D. R. (1972). Regression models and life-tables. Journal of the Royal Statistical Society: Series B (Methodological), 34(2), 187–220. [3] Harrell, F. E., Jr., Califf, R. M., Pryor, D. B., Lee, K. L., & Rosati, R. A. (1982). Evaluating the yield of medical tests. JAMA, 247(18), 2543–2546. [4] He, X., Wang, L., & Chen, Y. (2022). Epidemiology and survival outcomes of patients with primary intraocular lymphoma. BMC Cancer, 22, 6183. [5] Huang, Z., Li, W., Xu, G., et al. (2025). A nomogram to predict cancer-specific survival of ocular melanoma. Frontiers in Medicine, 12, 1638733. [6] Ishwaran, H., Kogalur, U. B., Blackstone, E. H., & Lauer, M. S. (2008). Random survival forests. Annals of Applied Statistics, 2(3), 841–860. [7] Kaplan, E. L., & Meier, P. (1958). Nonparametric estimation from incomplete observations. Journal of the American Statistical Association, 53(282), 457–481. [8] Stålhammar, G. (2022). Long-term relative survival in uveal melanoma: A meta-analysis. Communications Medicine, 2, 82. [9] Uno, H., Cai, T., Pencina, M. J., D’Agostino, R. B., Sr., & Wei, L. J. (2011). On the C-statistics for evaluating overall adequacy of risk-prediction procedures with censored survival data. Statistics in Medicine, 30(10), 1105–1117. [10] Witzenhausen, R., Damato, B., Charteris, D., et al. (2024). Survival benefit of primary tumor treatment in uveal melanoma. Cancers (Basel), 16(11), 638514.

Copyright

Copyright © 2025 Rohit Bhuse, Shaurya Sawant. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.