American Association for Cancer Research
10780432ccr183335-sup-210318_4_supp_5425927_pr76sg.docx (812.92 kB)

Supplementary Data from Early Modeled Longitudinal CA-125 Kinetics and Survival of Ovarian Cancer Patients: A GINECO AGO MRC CTU Study

Download (812.92 kB)
journal contribution
posted on 2023-03-31, 21:07 authored by Olivier Colomban, Michel Tod, Alexandra Leary, Isabelle Ray-Coquard, Alain Lortholary, Anne Claire Hardy-Bessard, Jacobus Pfisterer, Andreas Du Bois, Christian Kurzeder, Alexander Burges, Julien Péron, Gilles Freyer, Benoit You

Supplementary Methods. Basic Population Modeling. /// Supplementary Table S1. Characteristics of the patients included in the landmark survival analysis. Supplementary Table S2. Typical parameter estimates from the final semi-mechanistic model in the learning set AGO-OVAR 9. Supplementary Table S3. Predictive value of KELIM terciles and other prognostic factors regarding progression-free survival (PFS) and overall survival (OS) in univariate and multivariate models in every trial considered separately. Supplementary Table S4. Covariates associated with progression-free survival (PFS-months) in the 3 datasets. Supplementary Table S5. Impact of KELIM terciles on PFS (progression-free survival) and OS (overall survival) adjusted on risk classes defined by Oza et al. in the ICON 7 trial. Supplementary Table S6. Univariate Weibull accelerated failure time model regarding overall survival (OS) on the pooled datasets. /// Supplementary Figure S1. Semi-mechanistic model. AMT is the amount of drug administered. C1 and C2 are the drug concentration in compartment 1 and 2 (Arbitrary Units). Supplementary Figure S2. Goodness-of-fit plot for the AGO-OVAR 9 learning set. Predicted versus observed concentrations are shown for (left) the population and for (right) individuals. Black line: identity line. Supplementary Figure S3. Visual predictive checks for the 3 datasets. Supplementary Figure S4. Predictive value of KELIM and other prognostic factors regarding progression free survival (PFS, left panel) and overall survival (OS, right panel) assessed using C-index. Supplementary Figure S5. Linear regression between overall survival hazard ratios and KELIM ratios in the 3 phase III trials. Supplementary Figure S6. Outcomes of overall survival analyses with KELIM adjusted for the treatment arms (ARM) for the three datasets (the learning set AGO-OVAR 9; the two validation sets AGO-OVAR7 and ICON-7).



Regarding cancer antigen 125 (CA-125) longitudinal kinetics during chemotherapy, the actual predictive value of the Gynecologic Cancer Intergroup (GCIG) CA-125 response criterion is questioned. The modeled CA-125 elimination rate constant KELIM exhibited higher prognostic value in patients with recurrent ovarian cancer enrolled in the CALYPSO trial. The objective was to validate the higher predictive and prognostic values of KELIM during first-line treatments. Data from three large phase III trials were analyzed: AGO OVAR 9 [learning set: carboplatin-paclitaxel (CP) ± gemcitabine; n = 1,288]; AGO OVAR 7 (validation set: CP ± topotecan; n = 192); and ICON7 (validation set: CP ± bevacizumab; n = 1,388). The CA-125 profiles were fit with a nonlinear mixed-effect model during the first 100 days, and the individual KELIM were calculated. KELIM prognostic and predictive values for survival were assessed against GCIG criterion and other prognostic factors in univariate/multivariate analyses. The GCIG CA-125 endpoint provided no meaningful predictive/prognostic information. C-index analyses confirmed the higher predictive value of KELIM compared with GCIG criterion for progression-free survival and overall survival (OS). KELIM provided reproducible prognostic information. Patients with favorable KELIM ≥ upper tercile (0.0711 per days) consistently experienced better OS, with HRs between 0.44 and 0.58 (e.g., median OS >65 months vs. <35 months). Modeled KELIM provides higher predictive and prognostic information based on CA-125 longitudinal kinetics compared with GCIG response criteria during first-line chemotherapy. Integration of this endpoint in guidelines may be considered. Individual KELIM and survival simulations can be calculated at Further assessment of the surrogate value of KELIM treatment–related variations in a GCIG meta-analysis is warranted.See related commentary by Maitland et al., p. 5182