Virtual Library
Start Your Search
Woo Yul Byun
Author of
-
+
P79 - Immunotherapy (Phase II/III Trials) - Immunotherapy Plus Chemotherapy (ID 256)
- Event: WCLC 2020
- Type: Posters
- Track: Immunotherapy (Phase II/III Trials)
- Presentations: 1
- Moderators:
- Coordinates: 1/28/2021, 00:00 - 00:00, ePoster Hall
-
+
P79.01 - TCR Sequencing to Identify Responders in Patients with Stage III NSCLC Treated with Atezolizumab with Chemoradiation (AFT-16) (ID 3201)
00:00 - 00:00 | Presenting Author(s): Woo Yul Byun
- Abstract
Introduction
Immune checkpoint therapies have significantly enhanced overall survival for non-small cell lung cancer (NSCLC), but whether a specific patient will respond to therapy remains difficult to predict. We propose an approach of using T cell receptor (TCR) sequencing to address this uncertainty by comparing the T cell immune repertoires of patients with partial response (PR) to the repertoires of patients with progressive disease (PD), based on RECIST criteria.
Methods
We conducted a multi-institutional phase II Alliance Foundation Trials study, AFT-16 (NCT03102242), to explore the safety and efficacy of atezolizumab before and after chemoradiation (CRT). 63 patients with stage IIIA/B NSCLC and performance status 0-1 were enrolled with the plan to receive 4 cycles of neoadjuvant atezolizumab (1200 mg IV q 21 days) with restaging scans after cycles 2 and 4. Non-progressing patients continued on to receive carboplatin and paclitaxel (C/P) weekly with 60 Gy RT in 30 fractions, followed by 2 cycles of C/P consolidation and up to 9 months of additional adjuvant atezolizumab therapy. Peripheral blood mononuclear cell samples were collected before and after neoadjuvant treatment, and genomic DNA was isolated (gDNA). Survey resolution sequencing of the TCR CDR3 region was conducted using the immunoSEQ v3 platform (Adaptive Biotechnologies, Seattle, WA). We focused on blood samples from pre-therapy and after 2 or 4 cycles of atezolizumab. The pre-treatment and post-treatment Simpson clonality, observed richness, productive fraction, and nucleotide to amino acid convergence were calculated for all patients (pre-treatment: 18 PR, 12 PD; post-treatment: 17 PR, 7 PD). Significance was determined using the Mann-Whitney U test. These variables before treatment were used to train a random forest classifier. 100 trees and all four variables were considered at non-leaf nodes. The number of trees and variables were optimized by asymptotically minimizing out-of-bag error. Clonotypes that were significantly expanded or contracted after treatment were identified using a beta-binomial differential abundance model for patients with paired samples (16 PR, 7 PD).
Results
There are no significant differences in the clonality, richness, and convergence between patients who demonstrated PR and PD before or after treatment. The productive fraction was significantly higher for patients demonstrating PD compared to PR before treatment (P = .02), and after treatment (P = .03). The random forest produced a 23.3% out-of-bag error (class errors: PR: 0.17, PD: 0.33; random classifier error: 30%) in predicting outcomes using pre-treatment variables. PR patients demonstrated an average of 3.63 contracted clonotypes while PD patients demonstrated an average of 0.29 (P = 0.046).
Conclusion
Preliminary evidence suggests that there are detectable differences in the immune repertoires of patients demonstrating PR versus PD and that the differences before treatment may be used for early outcome prediction. Current shortcomings include the absence of tumor TCR sequencing, small numbers of patients, and low sequencing resolution that limit predictive machine learning. Future analysis will ideally include tumor TCR sequencing, comparisons of immune repertoires of additional patients, and “deep” sequencing data. We will also correlate TCR sequencing with clinical outcomes once these data are mature.
Support: Alliance Foundation Trials; Genentech. https://acknowledgments.alliancefound.org
