1240MO - Quantitative radiomics for the detection of symptomatic pneumonitis following chemoradiotherapy in patients with stage III unresectable NSCLC

Background

Pneumonitis is a serious common complication of radiotherapy and immunotherapy. It was reported in 33.9% (Grade [Gr] ≥2: 19.8%) and 17.9% (Gr ≥2: 11.7%) of patients (pts) with Stage III unresectable NSCLC who received consolidation durvalumab after concurrent chemoradiotherapy in clinical trial (PACIFIC; NCT02125461) and real-world (PACIFIC-R; NCT03798535) settings. Accurate radiomic analyses may enable early detection and treatment of symptomatic (i.e., Gr ≥2) pneumonitis (SP), which is crucial to optimal patient care. Here we assess the utility of radiomics in SP identification.

Methods

538 pts from the PACIFIC trial were analysed; 97 (18%) had SP and 441 (82%) did not have pneumonitis or had an asymptomatic event. A proprietary radiomic assessment with an auto segmentation model was applied to CT images to identify 4 abnormality patterns (honeycombing, reticulations, ground-glass opacities, and consolidation) and quantify other radiomic features. Regularised logistic regression models were developed and evaluated for discriminative power to identify SP. The first was trained with radiomic data; the second was trained with radiomic and clinical data. Reported odds ratios (ORs) are based on 10-repeat 5-fold cross validation; ORs >1 indicate positive association with SP.

Results

The first model was able to identify SP (AUC [95% CI]: 0.74 [0.72, 0.75]) using reticulation volume normalised by total lung capacity (TLC; OR [95% CI]: 1.58 [1.56, 1.61]) and lung volume (0.60 [0.59, 0.61]). Addition of clinical data enhanced the second model (AUC [95% CI]: 0.78 [0.77, 0.80]), which identified SP using reticulation volume normalized by TLC (OR [95% CI]: 1.83 [1.80, 1.86]), durvalumab treatment (2.55 [2.45, 2.65]), Asian race (3.11 [2.98, 3.23]), neutrophil count (1.97 [1.94, 2.01]), etoposide treatment (1.49 [1.43, 1.56]), Stage IIIA disease (1.77 [1.71, 1.83]), and baseline smoking status (0.56 [0.54, 0.57]). ORs from final models will be reported.

Conclusions

Models based on radiomic features and clinical variables can predict SP development. With refinement, such techniques may enable earlier diagnosis of SP in pts with NSCLC and other cancers. Validation with additional data sets is ongoing.

Clinical trial identification

Editorial acknowledgement

Medical writing support for the development of this abstract, under the direction of the authors, was provided by Eric Exner, MD, PhD, of Ashfield MedComms (New York, NY, USA), an Inizio company, and was funded by AstraZeneca.

Legal entity responsible for the study

AstraZeneca.

Funding

AstraZeneca.

Disclosure

V.D. Haakensen: Financial Interests, Personal, Advisory Board: AstraZeneca, BMS, Novartis; Financial Interests, Personal, Local PI: AstraZeneca, Lilly, MSD, Novartis; Financial Interests, Speaker, Consultant, Advisor: AstraZeneca, Pfizer, BMS, MSD, Janssen. J. Bar: Financial Interests, Personal, Advisory Board: AbbVie, Amgen, AstraZeneca, Bayer, BMS, Causalis, MSD, Merck Serono, Novartis, Roche, Takeda; Financial Interests, Personal, Principal Investigator: AbbVie, Amgen, AstraZeneca, Bayer, MSD, Roche, Takeda; Other, Institutional, Research Funding: Immunai, OncoHost, MSD, AstraZeneca. F. Belmans, A. Corsi, L. Libert: Financial Interests, Personal, Full or part-time Employment: Radiomics.bio. M. Flechet: Financial Interests, Personal, Full or part-time Employment: Radiomics.bio; Financial Interests, Personal, Project Lead: Radiomics.bio. C. C. Meca: Financial Interests, Personal and Institutional, Full or part-time Employment: Radiomics.bio; Financial Interests, Personal and Institutional, Leadership Role: Radiomics.bio; Financial Interests, Personal and Institutional, Ownership Interest: Radiomics.bio; Financial Interests, Personal and Institutional, Stocks or ownership: Radiomics.bio; Financial Interests, Personal and Institutional, Stocks/Shares: Radiomics.bio. N. Tsoutzidis: Financial Interests, Personal, Full or part-time Employment: Radiomics.bio; Financial Interests, Personal, Ownership Interest: Radiomics.bio. P. Chander: Financial Interests, Personal, Full or part-time Employment: AstraZeneca; Financial Interests, Personal, Stocks/Shares: AstraZeneca. K.A. Patwardhan: Financial Interests, Personal, Full or part-time Employment: AstraZeneca. J. Faria: Financial Interests, Personal, Full or part-time Employment: AstraZeneca; Financial Interests, Personal, Stocks or ownership: AstraZeneca. D. De Ruysscher: Financial Interests, Institutional, Advisory Board: AstraZeneca, BMS, Philips; Financial Interests, Institutional, Coordinating PI: AstraZeneca, BMS, Beigene; Financial Interests, Institutional, Funding: AstraZeneca, BMS, Philips, Beigene, Eli-Lilly, Olink; Financial Interests, Institutional, Research Funding: AstraZeneca, BMS, Philips, Beigene, Eli-Lilly, Olink. All other authors have declared no conflicts of interest.