Oops, you're using an old version of your browser so some of the features on this page may not be displaying properly.

MINIMAL Requirements: Google Chrome 24+Mozilla Firefox 20+Internet Explorer 11Opera 15–18Apple Safari 7SeaMonkey 2.15-2.23

Machine Learning-Based Model Shows Promise for Indicating Response to Immune Checkpoint Inhibition

Not all eligible patients respond to treatment with an immune checkpoint inhibitor
09 Dec 2020
Immunotherapy;  Cancer Intelligence (eHealth, Telehealth Technology, BIG Data)

A machine learning-based model using clinical and patient-reported data from patients with various cancers may be an accurate tool for prediction of the objective response rate (ORR) following treatment with immune checkpoint inhibitors (ICIs), according to findings presented at the ESMO Immuno-Oncology Virtual Congress 2020, held from 9 to 12 December 2020.

In this study (NCT3928938) Sanna Iivanainen of the Oncology and Radiotherapy Department, Oulu University Hospital in Oulu, Finland and colleagues investigated whether a machine learning model that combines clinical and patient-reported data could be used to predict patient response following treatment with ICI.

The investigators constructed a machine learning-based prediction model for the ORR using data of 31 patients receiving ICI therapies for various advanced cancers. Of the prospectively collected data sources used as inputs for the model, 63 were clinician-assessed treatment responses, and others included reported immune-related adverse events according to Common Terminology Criteria for Adverse Events (CTCAE) and laboratory measurements consisting of 9 values. Patient-reported symptom data including 18 monitored symptoms was collected using the Kaiku Health digital platform. The model also included the time from treatment initiation, as well as the patients’ age and sex. The laboratory values and patient-reported symptoms, were expressed as changes from baseline, and were linked to the treatment responses.

The prediction model for ORR was built using gradient a boosting technique (XGBoost). Prediction performance for unseen samples was evaluated using leave-one-out cross-validation (LOOCV), which trained and tested 63 models, each time iteratively leaving one sample out as a test set. The LOOCV prediction performance was evaluated with accuracy measures including the area under curve (AUC), F1 score, and Matthew’s correlation coefficient (MCC).

The model provided an ORR prediction following ICI therapy

The machine learning-based model was able to predict the ORR in patients undergoing ICI treatment for various cancers using XGBoost LOOCV performance with 75% accuracy, AUC 0.71, F1 score  = 0.58, and 0.40 MCC.


Non-zero feature importances of ORR prediction model trained with all available samples. The displayed importances depict the relative average improvement in prediction accuracy across all 100 trees in the model where a certain feature is utilised. The importances should be considered as relative to each other.

© Sanna Iivanainen.


According to the authors, this study demonstrated that it is possible to predict the ORR for patients undergoing ICI therapies with a machine learning model combining clinical, routine laboratory, and patient-reported data even with a limited size cohort.

They concluded that these promising results indicate that machine learning-based approaches in treatment response prediction warrant further investigation and validation with a larger dataset.

No external funding was disclosed for this study.


38P – Iivanainen S, Ekstrom J, Virtanen H, et al. Predicting objective response rate (ORR) in immune checkpoint inhibitor (ICI) therapies with machine learning (ML) by combining clinical and patient-reported data. ESMO Immuno-Oncology Virtual Congress 2020 (9-12 December).

This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.

For more detailed information on the cookies we use, please check our Privacy Policy.

Customise settings
  • Necessary cookies enable core functionality. The website cannot function properly without these cookies, and you can only disable them by changing your browser preferences.