||1R01CA249127-01 Interpret this number
||Kaiser Foundation Research Institute
||Leveraging Machine Learning to Improve Risk Prediction for Chemotherapy Induced Neuropathy
Chemotherapy-induced peripheral neuropathy (CIPN) affects more than two-thirds of adults with
invasive cancer who receive select adjuvant chemotherapies (e.g., taxanes, platinum analogs).
Severe CIPN symptoms can lead to chemotherapy dose reductions, treatment delays, or
changes in treatment regimens; thereby affecting the potential curative effects of chemotherapy.
For some patients, CIPN symptoms can persist over time, contributing to lower quality of life.
Little is known about risk factors for CIPN. Chemotoxicity risk scores have been developed
and evaluated for use among elderly patients receiving chemotherapy. However, these tools
generally report moderate predictive accuracy (60%-70%), small sample sizes, and short-term
follow up. We are aware of no publicly available, validated risk models to assess risk of severe
and chronic CIPN among diverse patients at risk for this potentially disabling side effect.
The goal of this proposal is to identify patients at risk for CIPN and to understand how
patients and provider interpret and use CIPN risk information in clinical decision-making.
Focusing on more than 8,500 insured adults (18+) diagnosed with invasive, stage I-III breast
and II-IIIA colorectal cancers (2013-2021) who received adjuvant chemotherapy treatment with
known risk for CIPN, we will develop and validate predictive models to quantify the risk of
severe CIPN and incident chronic CIPN and assess how CIPN risk information might be used to
inform clinical decision-making about cancer treatment and survivorship care planning.
We hypothesize that CIPN risk is a high priority for patients in thinking about treatment
choice and survivorship care planning. In addition, we hypothesize that the relative importance
of CIPN risk for patient and provider decision-making will vary by patient characteristics (e.g.,
age, cancer stage). We anticipate that the risk of severe and chronic CIPN can be predicted
with a high degree of accuracy using electronic health records and machine learning methods.
The study team has significant and complementary expertise in health services research,
biostatistics and predictive modeling, oncology practice, cancer epidemiology,
pharmacotherapy, drug safety and the patient care experience. To our knowledge, this will be
one of the first studies to develop and validate a CIPN predictive model that can be used by
oncology teams to inform treatment and care planning decisions and improve patient-valued
outcomes. Translation and replication of the findings will be catalyzed through publication in
peer-reviewed journals and the development and distribution of free software to facilitate testing
and adaptation of the resulting risk models across diverse systems of care.