Belén P.Solans (1, 2), Iñaki F. Trocóniz (1, 2)
(1) Pharmacometrics and systems pharmacology, Department of Pharmacy and Pharmaceutical Technology, School of Pharmacy, University of Navarra, Pamplona, Spain; (2) IdiSNA Navarra Institute for Health Research, Pamplona, Spain
Objectives: To develop a population K-PD model for tumour shrinkage effects of first line treatment in patients with locally advanced gastric cancer patients from whom scarce ordered categorical were obtained.
Methods: Data were obtained from 115 patients with locally advanced gastric cancer treated at the University Clinic of Navarra. Patients were divided in two protocols, one consisting on the administration of 3-4 cycles of neoadjuvant chemotherapy (CT) followed by neoadjuvant radiotherapy (CRT) (protocol 1); and the other consisting only in neoadjuvant CT and surgery (protocol 2). Surgery was scheduled 4 to 6 weeks after the end of the neoadjuvant treatment. Tumor size (CT scans) was measured at diagnosis and once or twice during treatment and prior to surgery; an additional tumour related measurement was obtained from biopsy (anatomy pathology) at surgery. Cancer stage of each individual was set using the TNM classification following the American Joint Committee on Cancer guidelines. From modelling purposes the TNM classification was reduced to 0 = partial response (downstage), 1 = stable disease (no change in the stage) and 2 = disease progression (upstage). At time of diagnosis it was assumed that all patients were on disease progression. Data were analysed under the population approach with NONMEN 7.3. Response measurements were treated as ordered categorical data using logistic regression and considering the contribution of first order markovian features.
Results: Significant treatment effects could be identified using the proposed analysis approach (other attempts did not succeed in extracting drug effects). Including time delays in drug effects did not improve data description. Similarly, data suggested that the markovian features, but not the tumor stage at diagnosis played a significant role in treatment effects.
Conclusions: Routine clinical data in general, and in oncology in particular, are sparse and scarce and represent a challenge from the modelling perspective. Herein we show that expressing TN response data in terms of down-, up-stage, and no change represents an alternative to characterize drug effects using logistic regression. Our analysis showed that markovian features were present in our data.
Reference: PAGE 25 (2016) Abstr 6022 [www.page-meeting.org/?abstract=6022]
Poster: Drug/Disease modeling - Oncology