A PK/PD mathematical model to forecast severe toxicities in pancreatic cancer patients treated with FOLFIRINOX regimen.

Abstract

e16774 Background: FOLFIRINOX regimen is the most efficient treatment in pancreatic adenocarcinoma. However, this triple-therapy causes significant and dose-limiting toxic effects leading to empirical dose-reduction, postponement of the forthcoming courses and sometimes treatment discontinuation. Dose-limiting toxicities with FOLFIRINOX are mostly oxaliplatin-induced peripheral neuropathy (OIPN) and irinotecan-induced diarrhoea (IID). The first aim of our study was to establish dose-concentration-toxicity relationships for these two main ordinal toxicities. Second, we aimed to perform in silico simulations to define the optimal FOLFIRINOX administration protocol (i.e., dosing, scheduling, sequencing) for a maximal benefit to risk ratio. Methods: We performed a retrospective pharmacometric analysis from clinical data collected in three French institutes. Patients had all confirmed diagnosis of PDAC treated with FOLFIRINOX as first line regimen. Demographic characteristics, toxicity data and dosage modifications were collected. Two mathematical models connecting dose – concentration – toxicity (PKPD Tox model) were developed using Monolix2019 to describe both OIPN and IID. Results: Data from 75 patients (36 females/39 males) treated with FOLFIRINOX regimen between 2015-2018 and representing 566 courses were collected. Median age was 65.4 years (range 29-78). The performance status at the start of treatment was 0-1 for 70% of patients. The most frequent location of the tumor was pancreatic head (50%). 91% of patients were subject to empirical dose reduction during their treatment. Oxaliplatin doses were decreased in 258 courses and stopped in 66 courses. Irinotecan doses were decreased in 257 courses and stopped in 33 courses. Overall, 130 IID and 198 OIPN events were observed. From these data, two different PKPD Tox models were developed. The individual concentrations of SN38 (active metabolite of irinotecan) and oxaliplatin were simulated with population PK models, either home-built or taken from the literature. Concentrations were translated into adequate exposure via an interface model developed by our group. Exposure was finally related to a pharmacodynamic model to predict probabilities assigned to each grade of IID or OIPN. Conclusions: These PKPD Tox models help to predict dose-limiting toxicities upon FOLFIRINOX and are useful to select the administration protocol of FOLFIRINOX minimizing the occurrence probabilities of treatment-related neuropathy and diarrhoea, while maintaining efficacy.