Industry insight

Predicting patient-reported symptoms for cancer patients undergoing immune checkpoint inhibitor (ICI) therapies using different measurement system than in prediction model training

Kaiku has continued its work among building machine learning (ML) based prediction models for immunotherapy-related symptom development. A year ago results from the first study were presented in the ESMO IO congress 2019 where it was showcased that ML based modeling of ePRO data on ICI treated cancer patients is feasible in predicting the onset and continuation of symptoms related to ICI toxicities. 

Now, in a study conducted together with Razvan Popescu & Tumor Zentrum Aarau from Switzerland it was proven that the previously built models generalize well to data collected using different symptom measurement system (PRO-CTCAE) in a different geographical area. The results are promising and Kaiku and Tumor Zentrum Aarau plan to continue the joint research.

The results of this study were recently published and presented in the well-known Swiss Oncology and Hematology (SOHC) 2020 congress. In the study, eight symptoms overlapped closely in training data and new dataset and were selected for model testing (shortness of breath, itching, cough, diarrhea, nausea, fatigue, rash and decreased appetite). Previously unseen test data consisted of 291 symptom reports by 21 patients treated with ICI therapies. Test data was presented to the models and the performance was evaluated. Results indicated that the models performed well in recognizing whether the symptom patient is experiencing will onset or continue.

Early detection of ICI toxicities could result in improved patient safety and better quality of life for patients.

Interested? Contact us here for more information. 



Get in touch with us

Interested in learning more about Kaiku Health? Send us a note