2022 World Conference on Lung Cancer (ePosters)-EP08.01-006. Using Real World data to build effective predictive machine learining models for NSCLC patients treated with immune-based therapy

EP08.01-006. Using Real World data to build effective predictive machine learining models for NSCLC patients treated with immune-based therapy

Back to course

Pdf Summary

This document discusses the use of machine learning models to predict the response to immune-based therapy in non-small-cell lung cancer (NSCLC) patients. Currently, PD-L1 expression is the primary biomarker used to predict response to immunotherapy, but it is not always accurate, as 40% of patients with high PD-L1 expression do not benefit from the therapy. Machine learning allows for the integration of patient and tumor data to improve the accuracy of prediction biomarkers. However, a major challenge is the lack of explainability of these models, making them difficult to interpret. Explainable Artificial Intelligence (XAI) is a tool that can help understand how machine learning models generate their results.<br /><br />The results of the study suggest that integrating real-world data through machine learning techniques can be a valuable tool in improving the prediction of response to immunotherapy in NSCLC patients. The best performing model achieved higher accuracy compared to using only the PD-L1 biomarker. The study found that certain features, such as NLR, Line of IO, ECOG, LDH, and PD-L1, were the most relevant in predicting outcomes. High values of NLR, Line of IO, ECOG, LDH, and low values of PD-L1 were associated with non-responders. The study also found that the combination of immunotherapy and chemotherapy benefited patients and predicted a positive response.<br /><br />The document provides a summary of the machine learning model's performance on the test dataset, showing its accuracy for different outcomes such as disease control rate, overall survival at 6 months, and objective response rate. The statistical analysis section provides information on the patient population used in the study, including the distribution of histology, sex, therapy type, and the percentage of responders/non-responders for each outcome.<br /><br />In conclusion, the integration of clinical, radiological, and hematological data through machine learning models can improve the prediction of response to immune-based therapy in patients with advanced NSCLC. The study highlights the importance of explainable AI in understanding how these models generate their predictions.

Asset Subtitle

Alessandro De Toma

Meta Tag

Speaker Alessandro De Toma

Topic Metastatic Non-small Cell Lung Cancer - Immunotherapy

Keywords

machine learning models

immune-based therapy

non-small-cell lung cancer

PD-L1 expression

predictive biomarkers

Explainable Artificial Intelligence

real-world data

NLR

combination therapy

advanced NSCLC