Katharina Och (1), Katharina Martha Götz (1), Amin Turki (2), Thorsten Lehr (1)
(1) Clinical Pharmacy, Saarland University, Saarbruecken, Germany, (2) Department of Bone Marrow Transplantation, West-German Cancer Center, University Hospital Essen, Germany
Introduction: Acute graft versus host disease (aGvHD) is an immunological reaction after allogeneic stem cell transplantation (HCT) where T-lymphocytes of the graft perceive tissues of the recipient as foreign. The induced immune response leads to damage in the affected organs (usually skin, gut or liver). Consequently, non-relapse mortality is increased and quality of life decreased [1]. To prevent aGvHD from happening, immunosuppressive drugs, primarily Cyclosporine A (CsA), are given.
Objectives: This project aimed to investigate the potential benefit of CsA in preventing aGvHD in adult patients after HCT, with a primary focus on onset and severity of aGvHD. Further objectives were to incorporate patient characteristics.
Methods: The model was build using retrospectively collected data derived from University Hospital Essen, Germany. All subjects received allogeneic HCT due to different hematological disorders. The observation period covers the first 100 days after transplantation, in which aGvHD may occur by definition. Patients were right censored (random type I censoring). Onset of aGvHD was the endpoint graded from 1 to 4 by the common used Seattle criteria, which takes severity and localization into account [2]. Plasma concentrations of CsA were measured during the therapeutic drug monitoring procedure. The time to event model was developed using non-linear-mixed-effects modelling technique, implemented in the software NONMEM® (version 7.3.0). Statistical evaluation was performed within the software R (version 3.4.1). To perform Visual Predictive Checks Pearl-speaks-NONMEM® (version 4.8.1) was used. Model selection criteria were defined as change in NONMEM objective function value >-3.84 and good fit of Kaplan-Meier plots of observed data vs. 97.5% confidence interval calculated from simulations of 200 replicates of the dataset.
Results: The dataset included 1353 patients with 1144 events of aGVHD grade 1-4. Whereas 1278 patients received CsA as immunosuppressive prophylaxis with 16295 recorded plasma levels, 12 patients received Tacrolimus (Tac) with 148 plasma levels and 63 received both CsA and Tac due to a switch from one to another with 846 resp. 641 plasma levels.
To describe the onset of aGvHD over time different hazard models (e.g. Gompertz, Weibull and proportional hazard) were tested. A non-monotonic function mimicking a one-compartmental model with six transit compartments described the hazard best. The course of grade 1-4 aGvHD hazard over time was found to be strongly correlated with the median monocytes in percentage over time and reaches its maximum at the time of the maximal monocytes percentage in plasma. In order to take the grade of aGvHD into account the model was extended by a proportional odds model with three transitions. CsA showed significant effect on the transition from grade 1 to 2. This effect were adequately described by sigmoidal Emax functions: Concentrations above 113 µg/L reduced the probability of the transition from grade 1 to grade 2. Moreover, aGvHD risk factors from literature were tested and the influence of stem cell source was confirmed.
Conclusions: The developed parametric time-to-categorical event model adequately described the time to the onset of aGvHD and its severity. Due to the pathophysiology of aGvHD, where antigen-presenting cells like monocytes play a key role in activating T-lymphocytes for inflammation, it is plausible that the non-monotonic hazard function correlates with monocytes. Furthermore, we found a quantitative relationship between CsA and aGvHD severity. Further research has to be done in order to improve the prediction and validate the model.
Funding: This work is part of the project “XplOit – Semantic Support for Predictive Modelling in Systems Medicine” which is funded by the German Federal Ministry of Education and Research (BMBF, grant id: 031L0027B).
References:
[1] Valkova V et al. Neoplasma (2016) 63, 743-51.
[2] Thomas ED et al. N Engl J Med (1975) 292, 895–902.
Reference: PAGE () Abstr 9518 [www.page-meeting.org/?abstract=9518]
Poster: Drug/Disease Modelling - Oncology