Abstract 1139P
Background
About 4-10% patients who receive immunotherapy are subjected to Immunotherapy-Induced Pneumonitis (IIP) which is a lethal complication that requires immediate interruption of immunotherapy as well as corticosteroid administration. Less than 10% of the patients also develop other types of pneumonitis (OP), (e.g. infection or radiation-induced). Making the differential diagnosis of pneumonitis (IIP vs. OP) is a challenging task due to similar radiological patterns and low incidence rates. This study was performed to develop and evaluate a prediction model using handcrafted radiomic features and their associated clinical metadata to estimate the risk of IIP in patients with stage IV Non-Small Cell Lung Cancer (NSCLC) who developed pneumonitis after immunotherapy.
Methods
Cause-specific pneumonitis events were identified among patients treated with immunotherapy in six centers in the Netherlands and Belgium from November 2017 to October 2020. Eight Regions of Interest (ROIs) (bounding box, segmented lungs, and spherical/cubical regions surrounding the inflammatory seed point) were examined to extract the most predictive radiomic features from chest computed tomography images obtained at the time of dyspnea manifestation. Models were internally validated in terms of discrimination, calibration, and decisional benefit.
Results
A total of 30 IIP and 42 OP events were identified. Cardiopulmonary comorbidities was the only predictive factor in the clinical model (OR=0.36, P=0.05). The best radiomic model reached an area under the receiver operating characteristic curve of 0.76 (95% CI: 0.68-0.88) and negative predictive value of 84% using the 75mm spherical ROI. HHH Firstorder Minimum (OR=8.74, P=0.016) and Original glszm Zone Entropy (OR=1.77, P=0.066) were included in the radiomic signature. Fair to good calibration and net benefits were achieved for the radiomic model across the entire range of probabilities.
Conclusions
Radiomic features discerned from the inflammation-specific spherical ROI are potentially useful to provide individually tailored information on having IIP via point-of-care decision support systems.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Bristol Myers Squibb.
Disclosure
All authors have declared no conflicts of interest.