Linda Wanika (1), Prof Mike Chappell (1), Dr Neil Evans (1), Dr Martin Johnson (2) and Helen Tomkinson (2)
(1) School of Engineering, University of Warwick, (2) Quantitative Clinical Pharmacology, Early Clinical Development, IMED Biotech Unit, Astrazeneca, Cambridge, UK
Introduction: Chronic Lower Respiratory Diseases (CLRDs) are a class of diseases that affect the trachea, bronchioles and alveoli, e.g. emphysema, bronchitis and interstitial lung disease (ILD). Patients who suffer from CLRD, often have difficulty breathing and some cancer patients may experience CLRD as an adverse event. An example of this is patients receiving Tyrosine Kinase Inhibitors (TKIs) which have been shown to induce ILD [1]. A plausible method for investigating the occurrence of CLRDs in cancer patients is through analysing biomarkers, routinely taken during a clinical trial. Lactate Dehydrogenase (LDH) is an enzyme found in most cells. LDH levels are usually increased when cellular damage is present and can be used as a biomarker for cell damage and apoptosis [2]. Some studies have concluded that an increase in LDH levels occurs during the occurrence of CLRD [3].
Objectives: This study aims to investigate the relationship between serum LDH levels and the occurrence of CLRD in cancer patients using data analytics and mechanistic modelling. A Cox’s Proportional Hazard model is used to investigate the occurrence of CLRD and a mechanistic PK/PD model to simulate the changes to serum LDH levels is used in order to assess whether LDH is a suitable predictive marker for CLRD.
Methods: Clinical trial data were obtained from Project Data Sphere, a clinical trial database platform [4]. Two clinical trials had 712 cancer patients (receiving the drug erlotinib as their treatment) with LDH measurements provided overtime, of which 207 patients experienced a CRLD related event. A Cox’s proportional hazard model was performed in Rstudio using Survival and Survminer [5-6]. The factors for the model were grouped baseline LDH levels and the %change between the baseline and the onset of a CLRD event. The PK/PD model is composed of a set of nonlinear parameterised ODEs, which simulated the change in LDH in response to erlotinib treatment. This was performed in Rstudio using RxODE [7].
Results: Compared to normal LDH levels (0-270 U/L), patients with a baseline value between 270-540 U/L were 1.5 times more likely to experience a CLRD (95% CI: 1.1, 2.1, p=0.019). Patients with a baseline LDH that was >540 U/L were 2.1 times more likely to experience a CLRD (95% CI: 1.2, 3.5, p=0.008). There was no significant risk associated with an increase in %LDH change. The model was able to predict a CLRD occurrence (10 days from the observed event time) for 15 patients.
Conclusions: This study confirms that high LDH levels are correlated against the occurrence of CLRDs and that the analysis of longitudinal LDH, (rather than just the baseline value), may be useful in investigating LDH as a possible predictor for CLRD. The model will be built upon by incorporating other analyses of LDH, such as slope measurements. This would also allow for the prediction of CLRD for patients who have a high baseline LDH value. The addition of covariates may also aid in the predictions of CLRD, as well the inclusion of more transit compartments in the PK/PD model to delay the LDH increase.
Acknowledgements: This project is funded by the EPSRC CASE award and Astrazeneca.
References:
[1] Shah R. (2016). Tyrosine Kinase Inhibitor-Induced Interstitial Lung Disease: Clinical Features, Diagnostic Challenges, and Therapeutic Dilemmas. Drug Safety. 39 (11) pp:1073-1091.
[2] Uchide N, et al. (2009). Lactate Dehydrogenase Leakage as a Marker for Apoptotic Cell Degradation Induced by Influenza Virus Infection in Human Fetal Membrane Cells. Intervirology. 52 (3) pp:164-173.
[3] Emad A and Reza Rezaian G. (1999) Lactate Dehydrogenase in Bronchoalveolar Lavage Fluid of Patients with Active Pulmonary Tuberculosis. Respiration. 66 (1) pp:41-45.
[4] Project Data Sphere (2019). URL: https://www.projectdatasphere.org/projectdatasphere/html/home. Accessed: 25/02/2019.
[5] Therneau T (2015). A Package for Survival Analysis in S_. version 2.38, URL:https://CRAN.R-project.org/package=survival. Accessed: 27/02/2019
[6] Alboukadel Kassambara and Marcin Kosinski (2018). survminer: Drawing Survival Curves using ‘ggplot2’. R package version 0.4.2. URL: https://CRAN.R-project.org/package=survminer. Accessed: 27/02/2019
[7]Wang W, et al. (2015). A Tutorial on RxODE: Simulating Differential Equation Pharmacometric Models in R. URL: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4728294. Accessed: 27/02/2019
Reference: PAGE 28 (2019) Abstr 9146 [www.page-meeting.org/?abstract=9146]
Poster: Drug/Disease Modelling - Other Topics