193O - Prognostic gene expression signature in chemotherapy treated patients from the MAGIC trial (193O)

Date 19 November 2017
Event ESMO Asia 2017 Congress
Session Proffered paper session 4
Topics Anti-Cancer Agents & Biologic Therapy
Gastric Cancer
Gastrointestinal Cancers
Translational Research
Presenter Elizabeth Smyth
Citation Annals of Oncology (2017) 28 (suppl_10): x57-x76. 10.1093/annonc/mdx660
Authors E.C. Smyth1, G. Nyamundanda2, D. Cunningham1, I.B. Tan3, E. Fontana2, C. Ragulan2, A.F. Okines1, S. Lin4, A. Wotherspoon5, M. Nankivell6, C. Peckitt7, N. Valeri2, R.E. Langley6, P. Tan8, A. Sadanandam2
  • 1Gastrointestinal Oncology, Royal Marsden Hospital NHS Foundation Trust, SW3 6JJ - London/GB
  • 2Molecular Pathology, The Institute of Cancer Research/Royal Marsden NHS Foundation Trust, SM2 5PT - Sutton/GB
  • 3Medical Oncology, National Cancer Centre Singapore, 169610 - Singapore/SG
  • 4Department Of Cancer Therapeutics And Stratified Oncology, Genome Institute of Singapore, Singapore/SG
  • 5Pathology, Royal Marsden Hospital NHS Foundation Trust, SW3 6JJ - London/GB
  • 6Medical Research Council Clinical Trials Unit, University College London, London/GB
  • 7Research & Development, Royal Marsden Hospital NHS Foundation Trust, London/GB
  • 8Cancer And Stem Cell Biology, Duke-NUS Medical School, Singapore/SG

Abstract

Background

Transcriptomics has defined novel molecular subgroups of gastroesophageal cancer (GC), however the prognostic value of these classifications has not been evaluated in the context of standard treatment. We hypothesised that gene expression on post-chemotherapy resection specimens from patients treated in the MAGIC trial could be used to create prognostic groups with different survival outcomes.

Methods

RNA was extracted from FFPE resections and analysed with the NanoString Technologies’ nCounter system. The gene panel included 200 genes associated with different GC characteristics. Penalised Cox regression was used to identify genes that predict overall survival (OS) followed by computing risk scores (GC-Assigner) for each patient using standard Cox regression. Finally, unsupervised analysis was used to cluster patients into GC-Assigner risk groups associated with OS.

Results

Gene expression data from 82 chemotherapy treated MAGIC trial patients were used to generate a 7 gene signature that predicts OS. Using GC-Assigner scores, three groups were defined; 3 year OS from surgery was 0% (95% 0 – 0%) for high risk patients, 40% (95% CI 27.0% - 64.0%) for intermediate risk patients and 80% (95% CI 63.8% - 99.8%) for low risk patients (p  0.05).

Conclusions

These data suggest that risk score and GC-Assigner groups are independent predictors of prognosis in GC patients treated with neoadjuvant chemotherapy in the MAGIC trial . As risk is assigned using post-treatment resection tissue which is less limited than diagnostic biopsies, pending our ongoing validation of this signature and these risk groups, GC-Assigner could be used as stratifier for future clinical trials evaluating personalised post-chemotherapy and resection treatment approaches for GC patients.

Clinical trial identification

ISRCTN93793971.

Legal entity responsible for the study

Medical Research Council Clinical Trials Unit at University College London

Funding

The TransMAGIC study was supported by CRUK grant (CRUKE/07/049).

Disclosure

E.C. Smyth: Honoraria for advisory role from Bristol Meier-Squibb, Five Prime Therapeutics and Gritstone Oncology. D. Cunningham: Research Funding from: Amgen (Inst); AstraZeneca (Inst); Bayer (Inst); Celgene (Inst); MedImmune (Inst); Merck Serono (Inst); Merrimack (Inst); Sanofi (Inst). A. Sadanandam: Licensor for a patent number PCT/IB2013/060416. Research funding from from Bristol-Myers Squibb. All other authors have declared no conflicts of interest.