526PD - Development of a predictive score using amphiregulin (AREG), epiregulin (EREG) and EGFR-FISH expression levels to determine treatment efficacy in mc...

Date 29 September 2012
Event ESMO Congress 2012
Session Gatrointestinal tumors, colorectal
Topics Colon Cancer
Rectal Cancer
Translational Research
Presenter Volker Heinemann
Authors V. Heinemann1, R. Laubender2, D.P. Modest3, A. Jung4, L. Fischer von Weikersthal5, U. Vehling-Kaiser6, C. Giessen3, T. Kirchner4, U. Mansmann2, S. Stintzing7
  • 1Dept. Of Medicine Iii, University of Munich, 81377 - Munich/DE
  • 2University Of Munich, Institute of Medical Informatics, Biostatistics, and Epidemiology, 81377 - München/DE
  • 3Department Of Medical Oncology And Comprehensive Cancer Center, University of Munich, 81377 - Munich/DE
  • 4Institute Of Pathology, Univsersity of Munich, 81377 - Munich/DE
  • 5Oncological Practice, Gesundheitszentrum St. Marien, Amberg/DE
  • 6Onkologische Schwerpunktpraxis, Onkologisches und Palliativmedizinisches Netzwerk Landshut, 84028 - Landshut/DE
  • 7Hämatologie Und Onkologie, University of Munich, 81377 - Munich/DE

Abstract

Background

We investigated the expression of the EGFR ligands amphiregulin (AREG) and epiregulin (EREG) as well as the amplification of the EGFR-gene in tumor specimens of mCRC patients (pts) treated first-line with anti-EGFR targeted cetuximab together with CAPOX or CAPIRI. All three factors have shown the capability to separate patient groups of different prognosis. Here we used a composite score to calculate response probability.

Methods

A total of 185 mCRC pts were randomized to cetuximab plus CAPIRI or plus CAPOX. The primary study endpoint was ORR. Age, gender, study arm, KRAS mutational status, BRAF mutational status, AREG, EREG and EGFR-FISH expression levels were used for multivariate logistic regression analysis.

Results

AREG and EGFR-FISH significantly correlated with ORR. These factors were used to create a prediction model for ORR by using logistic regression model. The linear predictor of this model is given by LP = -7.63 + (0.35*log[AREG]) + (6.58*[EGFR-FISH]) which can be used to calculate the probability of ORR by exp(LP) / [1 + exp(LP)]. The discriminatory ability of this model was assessed by ROC analyses were the optimal cut-off of the linear predictor discriminating best between responders and non-responders with respect to ORR was identified by using the Youden index and is given by -0.095. The area under the ROC curve (AUC) is 0.79 (95% confidence interval: 0.63, 0.90). Using this formula, response and survival times with a cut-off of 50% of ORR probability were calculated. ORR was significantly higher in the positive cohort and reached 81% vs. 42% in the negative cohort (p= 0.009). This led to significantly longer survival times in the positive cohort, with regard to PFS (8.6 mo vs 4.6 mo; HR 0.44; p = 0.003) and OS (38.4 mo vs 17.2 mo; HR 0.37; p= 0.001).

Conclusion

In this retrospective and exploratory analysis it was possible to predict ORR probability with the help of molecular markers. There is an acceptable discriminatory performance as measured by the AUC for predicting ORR by AREG and EGFR-FISH Hence, both molecular markers might be promising candidates for predicting ORR under cetuximab-based treatment regimens and prospective evaluation of these markers is needed for replicating and validating our findings.

Disclosure

V. Heinemann: Honoraria and travel support: Merck, Roche, Amgen.

R. Laubender: Travel support: Merck.

D.P. Modest: Honoraria and Travel Support: Amgen, Roche.

A. Jung: Honoraria and travel Support: Amgen, Roche.

C. Giessen: travel Support: Merck, Roche.

T. Kirchner: Research Grant, Honoraria and Travel Support: Merck, Amgen, Roche.

U. Mansmann: Honoraia and Travel support: Merck, Amgen.

S. Stintzing: Honoraria and travel support: Merck, Amgen, Roche.

All other authors have declared no conflicts of interest.