2363 - Investigation of the 'New-PDL1' Delta Radiomic Signature for Identifying Chemo-Radiotherapy Patients Who May Benefit from Consolidation Immunotherapy
Presenter(s)
A. R. Filippi1, J. Saddi2, V. Bartolomeo1, F. Agustoni2, G. Galli2, P. Borghetti3, S. La Mattina2, G. Facheris4, D. Cortinovis5, S. Arcangeli6, G. Piperno7, C. Aristei8, C. Bortolotto9, P. Pedrazzoli2, N. Tsoutzidis10, A. Corsi10, S. Ghezzo10, C. Meca10, and L. Preda11; 1Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy, 2Fondazione IRCCS Policlinico San Matteo, Pavia, Italy, 3Radiation Oncology Department, ASST Spedali Civili di Brescia - Brescia University, Brescia, Italy, 4Istituto del Radio, Brescia I-25100, Italy, 5Fondazione IRCCS San Gerardo dei Tintori, Monza, Italy, 6University of Milan Bicocca - School of Medicine and Surgery, Milan, Italy, 7European Institute of Oncology, Milan, Italy, 8Radiation Oncology Section, Department of Surgical and Biomedical Sciences, University of Perugia, Perugia General Hospital, Perugia, Italy, 9Fondazione IRCCS Policlinico San Matteo and University of Pavia, Pavia, Italy, 10Radiomics.bio, Liegi, Belgium, 11Department of Clinical-Surgical, Diagnostic and Pediatric Sciences, University of Pavia, Pavia, Italy
Purpose/Objective(s): Blue Sky Radiomics is a multicenter observational study conducted in Italy that investigates the prognostic and predictive role of radiomics in patients with stage III, unresectable, PD-L1+ non-small cell lung cancer (NSCLC) who are treated with chemoradiotherapy (CRT) followed by durvalumab maintenance therapy.
Materials/Methods: Eighty-five patients (N=85) from five Italian hospitals were enrolled. Clinical data, including histology, the date of immunotherapy initiation, and the date of first progression, were collected alongside contrast-enhanced CT scans at diagnosis (T0) and after CRT (T1). Patients were categorized as responders or non-responders based on Progression-Free Survival (PFS), utilizing a six-month threshold from the start of immunotherapy. The primary tumor (T) at both time points (T0 and T1) was delineated using a method that combined automatic thresholding with manual corrections to ensure accurate application of the radiomic signature. We conducted imaging quality assessments and extracted delta radiomic features to evaluate primary tumor evolution between T0 and T1. The ‘New-PDL1’ signature was used to predict patient response to immunotherapy.
Results: Among the 85 enrolled patients, only 63 had evaluable, high-quality images. The Kaplan-Meier curve analysis demonstrated that the ‘New-PDL1’ signature effectively stratified patients into responders and non-responders, revealing distinct PFS outcomes. The model achieved an overall accuracy of 74.6%, correctly predicting 47 out of 63 cases. While it displayed high sensitivity (84%), accurately classifying 42 out of 50 responders, its specificity was lower (38.5%), correctly identifying only 5 out of 13 non-responders. An additional investigation into the radiomic features indicated that several delta radiomic features associated with the texture of the primary tumor were statistically significant between responders and non-responders.
Conclusion: By utilizing a robust imaging and delineation quality assessment methodology, we demonstrate the predictive power of the ‘New-PDL1’ radiomic signature in identifying patients who may benefit from immunotherapy. However, the model's limited capacity to classify non-responders suggests a potential mismatch with its training data, which used images from T0 (pre-immunotherapy) to T1 (first follow-up post-immunotherapy), while the test data (screening/after CRT) may present different characteristics. The high sensitivity of the signature may indicate a shared radiomic pattern between responders to immunotherapy and radiotherapy. In contrast, biological differences in non-responders could confuse the model, resulting in lower specificity. Clinical Trial number: NCT04364776.