Abstract
Comprehensive genomic and transcriptomic profiling (CGTP) to predict pembrolizumab (P) benefit in patients (pts) with advanced solid tumors (STs)
Journal of clinical oncology, Vol.39(15_suppl), pp.2609-2609
05/20/2021
DOI: 10.1200/JCO.2021.39.15_suppl.2609
Abstract
Abstract only
2609
Background: P is approved in many ST types, however predictive biomarkers and the proportion of pts who benefit vary widely. Biomarkers beyond PD-L1 immunohistochemistry and comprehensive genomic profiling (CGP) based tumor mutation burden (TMB) may improve benefit prediction. We determined if treatment data and CGTP collected in an ongoing observational trial (NCT03061305) could predict pan-ST P benefit. Methods: Eligible advanced ST pts had QC-passing TMB and expression data from multiplex PCR based tissue CGTP on FFPE tissue (StrataNGS and an investigational test) and documented P treatment > 1 month. Real-world time to next treatment (TTNT) was defined as time in months from therapy start to new therapy start (after stopping initial therapy) or death. TMB and gene expression biomarker association with P TTNT was evaluated. Backward stepwise regression was performed to fit a multivariate Cox proportional hazards model; pts were assigned to four score groups (IRS 1-4) based on overlapping TTNT curves from 8 equal bins. P TTNT were compared between IRS groups by log-rank test. A chemotherapy (C) comparator cohort was established from C TTNT for pts in this cohort. Results were stratified by ST type, P mono vs. C combo, and TMB status. Results: 610 pts (254 [41.6%] NSCLC; 356 [58.4%] from 23 other ST types) with CGTP and P treatment were identified; P TTNT was highly correlated to overall survival (n=146; Pearsons r
2
=0.75). By univariate analysis of TMB and 9 expression biomarkers, TMB, two independent PD-L1 expression amplicons, and PD-L2 expression were significantly associated with P TTNT (all p ≤ 0.002). The most significant multivariate model included 5 variables, with 1) increasing TMB, PD-L1, and PD-L2, and 2) decreasing TOP2A (proliferation) and GZMA as P TTNT predictors. Median P TTNT, but not C TTNT (345 courses from 254 pts), differed significantly by IRS group (Table). Median P TTNT by IRS group did not significantly differ by non-small cell lung vs. other ST type or P mono vs. C combo (both p > 0.05); excluding TMB-high patients, median P TTNT was still significantly longer in IRS groups 3/4 vs. 1/2 (p = 5.0e-4). Across 19,623 total evaluable pts in NCT03061305, 12.2% were in IRS groups 3/4 and outside of P approved ST types/TMB-low. Conclusions: CGTP in an observational trial cohort demonstrated that TMB, PD-L1 and PD-L2 independently predicted pan-ST P benefit as assessed by OS-validated TTNT. A multivariate CGTP signature predicted P benefit relative to C across ST types. If further validated, such a signature may enable improved P benefit prediction. P versus C TTNT by IRS group. Clinical trial information: NCT03061305. [Table: see text]
Details
- Title: Subtitle
- Comprehensive genomic and transcriptomic profiling (CGTP) to predict pembrolizumab (P) benefit in patients (pts) with advanced solid tumors (STs)
- Creators
- Dan Rhodes - Strata Oncology, Ann Arbor, NCDaniel H Hovelson - Strata Oncology, Ann Arbor, MIMalek M. Safa - Kettering Medical CenterMark E. Burkard - University of Wisconsin Carbone Cancer CenterEddy Shih-Hsin Yang - University of Alabama at BirminghamWilliam Jeffery Edenfield - Greenville Memorial HospitalTravis Reeder - Strata Oncology, Ann Arbor, MIHana Vakil - Strata Oncology, Ann Arbor, MIKat Kwiatkowski - Strata Oncology, Ann Arbor, MIKhalis Mitchell - Strata Oncology, Ann Arbor, MIBryan Johnson - Strata Oncology, Ann Arbor, MIScott A. Tomlins - Strata Oncology, Ann Arbor, MI
- Resource Type
- Abstract
- Publication Details
- Journal of clinical oncology, Vol.39(15_suppl), pp.2609-2609
- Publisher
- LIPPINCOTT WILLIAMS & WILKINS; PHILADELPHIA
- DOI
- 10.1200/JCO.2021.39.15_suppl.2609
- ISSN
- 0732-183X
- eISSN
- 1527-7755
- Grant note
Strata Oncology
- Language
- English
- Date published
- 05/20/2021
- Academic Unit
- Internal Medicine
- Record Identifier
- 9984701251702771
Metrics
1 Record Views