Electronic patient-reported outcomes (ePROs) and machine learning (ML) in predicting the presence and onset of immune-related adverse events (irAEs) of immune checkpoint inhibitor (ICI) therapies
| Autor: | Vesa Kataja, Jussi Koivunen, Henri Virtanen, Jussi Ekström, Sanna Iivanainen |
|---|---|
| Rok vydání: | 2020 |
| Předmět: | |
| Zdroj: | Journal of Clinical Oncology. 38:e14058-e14058 |
| ISSN: | 1527-7755 0732-183X |
| DOI: | 10.1200/jco.2020.38.15_suppl.e14058 |
| Popis: | e14058 Background: ICIs have introduced novel irAEs, arising from various organ systems without strong timely dependency on initiation and discontinuation of the therapy. Early detection of the irAEs could result in improved safety profile of the treatment and better quality of life for patients. Symptom data collected by ePROs could be used as an input for ML based prediction models for early detection of irAEs. Methods: The utilized dataset consisted of two data sources. The first dataset consisted of 16 540 reported symptoms from 33 ICI-treated cancer patients, including 18 monitored symptoms collected using Kaiku Health digital platform. The second dataset included prospectively collected irAE data, including initiation and end dates, CTCAE class, and severity of 26 irAEs (the longest irAE lasted 799 days, and the shortest two days while median duration was 61 days). Two ML models were built using extreme gradient boosting, a well-known classification algorithm. Using the ePRO data, the first model was trained to detect the presence and the second model to detect the onset (0-21 days prior to diagnosis) of irAEs. The dataset was split into training (70 % of the data) and test sets (30 % of the data) by random allocation. The test set was left out from the model training and tuning, and was used only to evaluate the model performance. Results: The model trained to predict the presence of irAEs had an excellent performance with the test dataset. The prediction of the irAE onset was more difficult, but the model performance was still at a very good level. The performance metrics for the ML models are presented in Table. Conclusions: Current study suggests that ML based prediction models, using ePRO data as input for the models, can predict the presence and onset of irAEs with high accuracy. Thus, it indicates that digital symptom monitoring combined with ML could enable the detection of irAEs in ICI-treated cancer patients. The results should be validated with a larger dataset from prospective clinical trials. Clinical trial information: NCT03928938. [Table: see text] |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|