Leveraging in vitro data from novel drug candidates to prioritize antibody combinations in autoimmune disease using a QSP model of IBD
Dominic G. Whittaker (1,3), Aijaz Shaliban (2), Mrittika Roy (2), Prakash Packirisamy (2), Maithreye Rengaswamy (2), Valeriu Damian (3), Anubha Gupta (1)
(1) Clinical Pharmacology Modelling and Simulation, GSK UK; (2) Vantage Research, LLC; (3) Systems Modelling and Translational Biology GSK PA, USA
Objectives:
For many autoimmune indications the treatment response has reached a ceiling effect and so novel combination treatment approaches may be required [1]. Given the numerous possibilities, early identification and prioritization of potential combinations is of utmost importance. Quantitative Systems Pharmacology (QSP) modelling can aid in these objectives by combining available mechanistic information on disease pathophysiology and mechanism of action of therapy with clinical outcomes. One of the challenges faced in QSP modelling while predicting the efficacy of new therapies and combinations is the lack of relevant, quantitative data on downstream effects, e.g. which cytokines are impacted and by how much.
For novel compounds, in vitro data on the pharmacodynamic effect are often available. However, these experiments are typically conducted under conditions where efficacy estimates may exceed normal physiological levels, making them unsuitable for direct model parameterization. We focused on addressing this issue using inflammatory bowel disease (IBD), a class of autoimmune diseases including Crohn’s disease (CD) and ulcerative colitis (UC) [2], as a case study, due to the abundance of available in vitro and clinical data for calibration and validation of a QSP model. We present here a method for deriving a scaling factor from in vitro data which allows such data to be leveraged directly in QSP model development and subsequent prediction of the clinical efficacy of novel combinations.
Methods:
- We adapted a published QSP model of IBD [3,4] that is calibrated to a range of therapies with respect to C-reactive protein and fecal calprotectin. We added a representation of a clinical outcome for UC, the Mayo score, which is modelled as an empirical score and was calibrated by assuming a correlation with various inflammation-related species. We also added relevant mechanisms to represent the mechanism of action of novel therapies of interest, resulting in an updated QSP model with representations of 20 cell types and 26 effectors.
- A Virtual Population (VPop) was created that was consistent with reported baseline patient characteristics and response to Adalimumab, Vedolizumab, and Ustekinumab (% remission and % response based on Mayo score).
- The final model was validated using the VEGA trial [1] - a published combination trial of Guselkumab (anti-IL23) and Golimumab (anti-TNFα) for UC.
- To implement novel therapeutics in the model, in vitro data were used which captured cytokine levels in healthy PBMCs in response to different doses of the novel drug candidates and anti-TNFα. A scaling factor was determined by comparing the reduction observed in TNFα in response to anti-TNFα in the in vitro data with the reduction observed in the clinical data in response to Adalimumab. This scaling factor was then applied to estimate the clinical reduction in cytokine levels for novel therapies.
- The calibrated and validated VPop was used to predict the remission and response percentages for novel combinations (anti-TNFα + novel monotherapy).
Results:
The calibrated QSP model VPop showed quantitative agreement with reported VEGA trial remission and response percentages (mean [95% CI]):
- Guselkumab monotherapy: remission (QSP: 10.2% [4.2-16.9] vs. Data: 10.1% [0.0-19.9]) and response (QSP: 34.2% [23.9-45.1] vs. Data: 40.7% [20.6-59.7])
- Golimumab monotherapy: remission (QSP: 7.6% [2.8-14.1] vs. Data: 11.2% [1.1-21.0]) and response (QSP: 15.9% [8.5-24.0] vs. Data: 27.2% [7.1-46.2])
- Combination therapy: remission (QSP: 18.9% [11.3-28.2] vs. Data: 25.6% [15.5-35.4]) and response (QSP: 58.9% [47.9-69.0] vs. Data: 49.2% [29.0-68.2])
The VPop was subsequently used to predict remission/response percentages for novel combination therapies across a range of doses, based on which the proposed combinations were ranked.
Conclusions:
We present the extension of a publicly available model of IBD to represent a commonly used measure of disease severity and calibrated and validated against several biologics of interest. Using contextual interpretation of the available in vitro data of novel drug candidates, the model was used to predict the clinical efficacy of proposed combinations of novel therapeutics with anti-TNFα. The method used here for the normalization of in vitro data can be used for other therapeutics as well to obtain more realistic predictions of clinical efficacy when other data is not available.
References:
[1] Feagan BG, Sands BE, Sandborn WJ, Germinaro M, Vetter M, Shao J, Sheng S, Johanns J, Panés J; VEGA Study Group. Guselkumab plus golimumab combination therapy versus guselkumab or golimumab monotherapy in patients with ulcerative colitis (VEGA): a randomised, double-blind, controlled, phase 2, proof-of-concept trial. Lancet Gastroenterol Hepatol. 2023 Apr;8(4):307-320.
[2] Eltantawy, N., El-Zayyadi, I.A.EH., Elberry, A.A. et al. A review article of inflammatory bowel disease treatment and pharmacogenomics. Beni-Suef Univ J Basic Appl Sci 12, 35 (2023).
[3] Rogers KV et al. A Dynamic Quantitative Systems Pharmacology Model of Inflammatory Bowel Disease: Part 1 – Model Framework, Clin Transl Sci (2021) 14, 239–248.
[4] Rogers KV et al. A Dynamic Quantitative Systems Pharmacology Model of Inflammatory Bowel Disease: Part 2 – Application to Current Therapies in Crohn’s Disease, Clin Transl Sci (2021) 14, 249–259.
