503PD - A genetic response profile to predict efficacy of adjuvant 5-FU in colon cancer

Date 27 September 2014
Event ESMO 2014
Session Gastrointestinal tumours, colorectal
Topics Colon and Rectal Cancer
Pathology/Molecular Biology
Basic Scientific Principles
Presenter Ida Buhl
Citation Annals of Oncology (2014) 25 (suppl_4): iv167-iv209. 10.1093/annonc/mdu333
Authors I.K. Buhl1, S. Gerster2, M. Delorenzi3, T. Jensen4, P.B. Jensen4, N. Brunner1, A. Hansen4, S. Knudsen5
  • 1Department Of Veterinary Disease Biology, Molecular Disease Biology, University of Copenhagen, 2100 - Copenhagen/DK
  • 2Bioinformatics Core Facility, Swiss Institute Bioinformatics (SIB), 1015 - Lausanne/CH
  • 3Bioinformatics Core, Swiss Institute of Bioinformatics, 1015 - Lausanne/CH
  • 4Headquarters, Medical Prognosis Institute, 2970 - Hoersholm/DK
  • 5Us Office, Medical Prognosis Institute, 85258 - Scottsdale/US



There is a clinical need for biomarkers of response to adjuvant 5-fluorouracil (5-FU) in colon cancer (CC). The current project seeks to validate a predictive 5-FU gene expression profile. A similar model has recently been validated with MD Anderson in three different clinical settings [1] and with researchers from AstraZeneca with fulvestrant [2].


The 5-FU profile consisted of in total 205 positively and negatively correlated genes mapped to 669 probe sets (on the ALMAC Colon DSA). This profile was tested using expression data obtained from a Colon DSA gene expression array from ALMAC applied on the PETACC-3 CC population [3] with formalin fixed paraffin embedded (FFPE) tissue from 636 stage III CC patients treated adjuvantly with 5-FU with or without irinotecan. It was also tested on ALMAC Colon DSA data obtained from FFPE tissue from 359 stage II CC patients who did not receive adjuvant treatment [4]. The analyses were performed using Cox proportional hazards (CPH) regression models and Kaplan-Meier curves. The logrank test was used to compare the survival distributions.


In the PETACC-3 cohort the 5-FU predictive profile showed a statistically significant association with overall survival (OS) (hazard ratio (HR) = 0.47 (0.34, 0.63), P = 7.4e − 07; binary scores). In the untreated cohort no differences were observed in OS (HR = 0.96 (0.67, 1.4), P = 0.849; binary scores). The statistical significance of the effect of the profile score (as continuous variable) adjusted for several relevant clinicopathological parameters including microsatellite instability (MSI) vs stability (MSS) and KRAS mutation status was confirmed in a multivariable CPH model on the PETACC-3 data (Table).

PETACC-3, HR estimates, variables with P < 0.05
5FU predict 1 unit = 1 iqr 0.69 0.58, 0.82
T stage T4 vs T3 2.03 1.4, 2.95
N stage N2 vs N1 2.01 1.45, 2.79
grade G34 vs G12 1.81 1.07, 3.07
KRAS mut vs wt 1.89 1.34, 2.67
MSI, MSI vs MSS 0.42 0.21, 0.81


Our data could suggest that the present 5-FU signature provides independent predictive information regarding benefit from adjuvant 5-FU in CC patients. [1] JNCI 2013;105:1284-1291 [2] PLoS One 2014;9:e87415 [3] JCO 2009;27:3117-3125 [4] JCO 2011;29:4620-4626.


I.K. Buhl: is affilated, but has no employment nor stock ownership in Medical Prognosis Institute (MPI). T. Jensen: P.B. Jensen, A. Hansen and S. Knudsen: has employment and stock ownership in Medical Prognosis Institute (MPI). All other authors have declared no conflicts of interest.