857PD - The predictive value of the CA-125 modeled kinetic parameter KELIM is validated in 3 independent datasets (AGO-OVAR 7 & 9; ICON 7 AGO/GINECO/GCIG t...

Date 08 October 2016
Event ESMO 2016 Congress
Session Gynaecological cancers
Topics Gynaecological Malignancies
Presenter Benoit MP You
Citation Annals of Oncology (2016) 27 (6): 296-312. 10.1093/annonc/mdw374
Authors B.M. You1, O. Colomban1, M. Tod1, I.L. Ray-Coquard2, A. Lortholary3, A.C. Hardy-Bessard3, A. Du Bois4, J. Huober5, W. Meier6, C. Kurzeder4, J. Pfisterer7
  • 1Emr Ucbl-hcl 3738, Université Claude Bernard Lyon 1, 69310 - Pierre Bénite/FR
  • 2Département D'oncologie Médicale Adulte, Centre Léon Bérard, 69008 - Lyon/FR
  • 3Oncologie, Centre Catherine de Sienne, Nantes/FR
  • 4Gynecology, Kliniken Essen Mitte Evang. Huyssens-Stiftung, Essen/DE
  • 5Gynecology, Ulm Medical University, Ulm/DE
  • 6Gynecology, Evangelisches Krankenhaus Düsseldorf, Düsseldorf/DE
  • 7Gynecology, Womens Cancer Center, Kiel/DE



Mathematical modeling can be used to analyze longitudinal CA125 kinetics, and predict treatment efficacy. The modeled CA-125 elimination parameter KELIM was a predictive factor of efficacy in CALYPSO trial (You et al. Gynecol Oncol 2013). The present study aims at validating the independent predictive value of KELIM in phase III trial datasets with different 1stline treatments.


Data from AGO-OVAR 7 (carboplatin-paclitaxel (CP) +/- topotecan; n = 1308); AGO-OVAR 9 (CP +/- gemcitabine; n = 1742) and ICON-7 trials (CP +/- bevacizumab; French dataset only, n= 196) were analyzed. The biggest AGO 9 dataset was used as a training set, while AGO 7 & ICON7 were used as validation sets. CA125 concentration-time profiles was fit with following parameters: tumor growth rate (BETA); CA 125 tumor production (KPROD); CA 125 elimination rate (KELIM) & treatment indirect effect (Emax relationships) “d[CA125]/dt = (KPROD* exp (BETA*t)) * (1 - (A/(A + A50))) – KELIM * [CA125]” where t is time. The predictive value of KELIM dichotomized by the median was tested regarding progression free survival (PFS) against other reported prognostic factors (stage; pathology; surgery/completeness if any; grade; arms; Rustin) using Cox-models.


Individual CA125 profiles were well described by the model in training and validation datasets, as validated by visual predictive checks. KELIM ( 0.0598) exhibited strong independent predictive value regarding PFS in training dataset (univariate: 24.0 vs 11.3 months, P 


The independent predictive value of KELIM was reproducible in large 3 datasets of ovarian cancer patients treated with different regimens. This may be a novel predictive factor, helpful for early selection of the best candidates during drug development.

Clinical trial identification

Not applicable

Legal entity responsible for the study

Benoit You




All authors have declared no conflicts of interest.