Forecasting patient-specific tumor response using patient-reported outcomes in non-small cell lung cancer

PurposePatient-reported outcomes (PROs) provide a quantitative measure of a patients quality of life, directly from the patient without external influence or interpretation. Prior studies have demonstrated correlations between individual PROs and cancer treatment response. However, this area of research is still highly understudied, and patient data often goes ignored. Our previous work has shown how changes in insomnia can be used to make binary decisions about a patients future volume response. Here, we expand upon that work to determine precisely when treatment progression will occur, providing an opportunity for clinicians to intervene sooner. Experimental DesignThis study analyzed PROs and tumor volume data collected from 80 NSCLC patients undergoing immunotherapy to determine how PRO dynamics could inform when volumetric treatment progression would occur. We calibrated the tumor growth inhibition (TGI) model to patient-specific tumor volume dynamics for all volume measurements using a leave-one-out cross-validation approach. Growth parameters were divided based on progression status and sampled depending on changes in patient-reported insomnia. A cutoff analysis was performed to determine the optimal cutoff for distinguishing between responders and non-responders. Predictions were made for the Nth patient and categorized using the cutoff. ResultsThis study demonstrated that incorporating patient-specific changes in insomnia with a mathematical model of volume changes can predict patient response with a 72.2% true positive rate and 71.3% overall accuracy, on average 6-8 weeks sooner. ConclusionUsing this innovative framework, we can predict precisely when progression occurs, giving clinicians the opportunity to intervene beforehand.