Machine Learning-Assisted Ensemble Analysis for the Prediction of Response to Neoadjuvant Chemotherapy in Locally Advanced Cervical Cancer

• Published in: Frontiers in oncology Vol. 12; p. 817250
• Main Authors: Huang, Yibao, Zhu, Qingqing, Xue, Liru, Zhu, Xiaoran, Chen, Yingying, Wu, Mingfu
• Format: Journal Article
• Language: English
• Published: Switzerland Frontiers Media S.A 29.03.2022
• Subjects: locally advanced cervical cancer; machine learning analysis; neoadjuvant chemotherapy; Oncology; pathology response; predictive model; pathology response; neoadjuvant chemotherapy; predictive model; locally advanced cervical cancer; machine learning analysis
• ISSN: 2234-943X; 2234-943X
• DOI: 10.3389/fonc.2022.817250

The clinical benefit of neoadjuvant chemotherapy (NACT) before concurrent chemoradiotherapy (CCRT) vs. adjuvant chemotherapy after CCRT is debated. Non-response to platinum-based NACT is a major contributor to poor prognosis, but there is currently no reliable method for predicting the response to NACT (rNACT) in patients with locally advanced cervical cancer (LACC). In this study we developed a machine learning (ML)-assisted model to accurately predict rNACT. We retrospectively analyzed data on 636 patients diagnosed with stage IB2 to IIA2 cervical cancer at our hospital between January 1, 2010 and December 1, 2020. Five ML-assisted models were developed from candidate clinical features using 2-step estimation methods. Receiver operating characteristic curve (ROC), clinical impact curve, and decision curve analyses were performed to evaluate the robustness and clinical applicability of each model. A total of 30 candidate variables were ultimately included in the rNACT prediction model. The areas under the ROC curve of models constructed using the random forest classifier (RFC), support vector machine, eXtreme gradient boosting, artificial neural network, and decision tree ranged from 0.682 to 0.847. The RFC model had the highest predictive accuracy, which was achieved by incorporating inflammatory factors such as platelet-to-lymphocyte ratio, neutrophil-to-lymphocyte ratio, neutrophil-to-albumin ratio, and lymphocyte-to-monocyte ratio. These results demonstrate that the ML-based prediction model developed using the RFC can be used to identify LACC patients who are likely to respond to rNACT, which can guide treatment selection and improve clinical outcomes.