Modeling Dose-Function Response and Toxicity Pathways in Non-Small Cell Lung Cancer Patients Undergoing Radiation Treatment

Abstract

Radiotherapy is currently the standard of care for treating NSCLC patients, and while technological advancements in the field continue to improve our ability to successfully treat these patients, serious side-effects, i.e. toxicities, may occur in response to the delivered radiation. In fact, Grade 2 RP {i.e. symptomatic, requiring medical intervention} has been reported to occur in 20-30% of NSCLC patients that receive RT, whereas Grade 5 RP {i.e. death directly related to radiation treatment} is estimated to occur in ~5% of NSCLC patients that undergo RT.

In an effort to better understand functional lung response to radiation, our group developed a novel method to model the patient-specific dose-function response using perfusion and ventilation SPECT intensity as a surrogate measure of lung function. Because a patient’s signature, i.e. patient-, treatment-, and disease-related factors, influences the dose-function response across all dose bins, there is an inherent correlation amongst data points contributed by each patient. To account for these interdependencies, a mixed-effects nonlinear regression model was implemented to allow for patient-specific parameters to be assigned to each patient’s dose-function response curve individually. Once each patient’s dose-function response was modeled, a population-level model was derived by averaging the patient-specific parameters to more accurately represent the expected dose-function response for a future arbitrarily selected patient. As such, we believe this patient-specific modeling approach can facilitate an enhanced characterization of personalized functional changes from a population-based estimate.

Furthermore, by measuring the dose delivered to functional lung categorizations in NSCLC patients undergoing RT, this thesis explicitly analyzed specific dose-function vulnerabilities that may lead to increased rates of toxicities and found that high dose to low-functioning lung was strongly associated with RILT incidence. Although surprising and contrary to the prevailing mantra, this result suggests that low-functioning regions of the lung, which are indicative of pulmonary dysfunction and possibly underlying disease, are susceptible to high dose and should not be disregarded in treatment planning. While it is generally accepted that the primary driving force of toxicity is dose to the normal lung, it is also known that pulmonary comorbidities can become exacerbated in response to radiation and have the potential to influence the incidence of severe forms of RILT. Consequently, by better understanding the mechanisms that cause functional damage and the various toxicity pathways, there is great potential to further mitigate RILT rates in NSCLC patients undergoing RT.

Based on these findings, a preliminary investigation regarding the utility of identifying and quantifying specific phenotypes of pulmonary disease prior to RT was performed. By utilizing PRM of high-resolution inspiration/expiration CT scans, pre-treatment voxelwise classifications that characterized lung parenchyma as normal, emphysema, small airways disease, or parenchymal disease were analyzed in a cohort of lung cancer patients to determine the expected distribution of each PRM category and to assess their correlation with RILT incidence. As a CT-based imaging technique, PRM offers significant upside due to its wide-availability and its capability to provide spatially-resolved estimates for various forms of pulmonary disease.

In summary, the aims of this thesis were to better understand the dose-function response in lung cancer patients during and after RT, identify functional lung targets that may be useful in mitigating toxicity incidence, and propose solutions to enhance personalized radiation treatment of NSCLC patients in an effort to optimize patient outcomes.