Prediction of tumour response induced by chemotherapy using modelling of CA-125 kinetics in recurrent ovarian cancer patients

Background: The main objective of the present study was to establish the relationships between CA-125 kinetics and tumour size changes during treatment. Methods: The data from the CALYPSO-randomised phase III trial, comparing two platinum-based regimens in recurrent ovarian cancer (ROC) patients, was randomly split into a `learning data set' to estimate model parameters and a ` validation data set' to validate model performances. A kinetic-pharmacodynamic semi-mechanistic model was built to describe tumour size and CA-125 kinetics during chemotherapy. The ability of the model to predict tumour response induced by chemotherapy, based on CA-125 values, was assessed. Results: Data from 535 ROC patients were used to model CA-125 kinetics and tumour size changes during the first 513 days after treatment initiation. Using the validated model, we could predict with accuracy the tumour size changes induced by chemotherapy based on the baseline imaging assessment and longitudinal CA-125 values (mean prediction error: 0.3%, mean absolute prediction error: 10.6%). Conclusions: Using a semi-mechanistic model, the dynamic relationships between tumour size changes and CA-125 kinetics induced by chemotherapy were established in ROC patients. A modelling approach allowed CA-125 to be assessed as a biomarker for tumour size dynamics, to predict treatment efficacy for research and clinical purposes.