Predictive Value of Machine Learning for Platinum Chemotherapy Responses in Ovarian Cancer: Systematic Review and...

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Ovarian cancer treatment often involves platinum-based chemotherapy.
• Predicting patient response to this therapy is crucial for personalized treatment strategies.
• Machine learning (ML) offers potential for predicting treatment outcomes, but its efficacy requires systematic evaluation.

Approach:

• A systematic literature review was conducted following PRISMA guidelines.
• Searched PubMed, Embase, Web of Science, and Cochrane databases for studies up to April 26, 2023.
• Evaluated 19 studies with 39 models using the Prediction Model Risk of Bias Assessment tool and performance metrics like C-index, sensitivity, and specificity.

Key Points:

• Logistic regression, Extreme Gradient Boosting, and Support Vector Machine were common ML methods.
• Pooled C-index was 0.806 in training and 0.831 in validation cohorts.
• Support Vector Machine demonstrated strong predictive performance (C-index 0.942 training, 0.879 validation). Pooled sensitivity and specificity were 0.890 and 0.790 respectively in the training cohort.

Conclusions:

• Machine learning models can effectively predict patient response to platinum-based chemotherapy for ovarian cancer.
• These findings support the use of ML in developing and refining clinical scoring systems for ovarian cancer treatment.
• ML-based prediction offers a valuable reference for future clinical decision-making and treatment optimization.