Abstract 3P
Background
Immune checkpoint inhibitor (ICI) related adverse events (irAE) can impact quality of life and necessitate the discontinuation of ICI treatment in melanoma patients. This study utilised machine learning (ML) and explainable artificial intelligence (AI) to predict irAEs in patients with melanoma as this could improve patient outcomes and support informed, shared decision-making between patients and clinicians.
Methods
455 datasets were included for patients initiated on ICIs between 2014-2024 in a single, large regional cancer in Kent, U.K. Six ML algorithms were developed using 80% of the data for training, and 20% for validation. Shapley Additive Explanations (SHAP) explainable artificial intelligence was used to interpret the irAE prediction model to improve the interpretation and transparency of the decision boundary.
Results
Of the 455 patients, 294 patients (65%) were treated with palliative intent, while 161 (35%) were treated with curative intent. 121 patients (27%) experienced irAEs with only 29 having a documented grade of irAE (grade 1-4) whilst the remaining 92 patients had unspecified irAE grades. The commonest class of irAE (CTCAE v5) was gastrointestinal (38, 31%), followed by skin (25, 21%) and subsequently endocrine (22, 18%). Of the 121 patients with irAEs, 36 (30%) required hospital admission as a consequence of their irAEs and 49 patients (40%) stopped ICI treatment as a consequence of their irAEs. Linear Regression and Gaussian Naïve Bayes predicted irAEs with the most accuracy, 92%. The model highlighted increasing age, female gender and exposure to combination therapy or pembrolizumab as predictors for irAE.
Conclusions
These ML algorithms accurately predicted irAEs in melanoma patients. Key predictors of irAE included increasing age, female gender and exposure to combination therapy or pembrolizumab. Future work will aim to externally validate these models and incorporate them into ICI drug development trials to prospectively predict which patients will develop irAEs. Our models can also be used as tools to support informed decision-making between clinicians and patients considering starting, stopping or continuing ICI therapy.
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Curenetics.
Disclosure
All authors have declared no conflicts of interest.