Modelling real-world tumor size dynamics based on electronic health records and image data in advanced melanoma patients receiving immunotherapy
Perrine Courlet (a,b)*, Daniel Abler (a,c)*, Pascal Girard (d), Alain Munafo (d), Monia Guidi (b,e), Chantal Csajka (b,f,g), Olivier Michielin (a), Michel Cuendet (a,h)◦, Nadia Terranova (d)◦ *co-first, equal contribution to the work ◦co-senior
(a) Precision Oncology Center, Department of Oncology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland. (b) Centre for Research and Innovation in Clinical Pharmaceutical Sciences, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland. (c) Institute of Information Systems, University of Applied Sciences Western Switzerland (HES-SO), Sierre, Switzerland. (d) Merck Institute of Pharmacometrics, Lausanne, Switzerland, an affiliate of Merck KGaA, Darmstadt, Germany. (e) Service of Clinical Pharmacology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland. (f) Institute of Pharmaceutical Sciences of Western Switzerland, University of Geneva, University of Lausanne, Geneva, Switzerland. (g) School of Pharmaceutical Sciences, University of Geneva, Geneva, Switzerland. (h) Swiss Institute of Bioinformatics, University of Lausanne, Lausanne, Switzerland.
Objectives:
Immune checkpoint inhibitors (ICIs) such as ipilimumab (CTLA4 inhibitor), nivolumab or pembrolizumab (PD1 inhibitors) have revolutionized cancer therapy with durable response in a subset of patients (39% five-year treatment-free survival [1]). Modeling of tumor size dynamics based on data from randomized controlled trials (RCTs) has been successfully reported in literature to allow the identification of sources of variability and the prediction of long-term outcomes in cancer patients receiving ICIs [2, 3]. While RCTs are generally considered to provide the highest level of evidence, retrospective analyses of real-world data (RWD) better reflect heterogeneous populations, diverse tumor phenotypes and treatment combinations. In this work, we extracted longitudinal RWD tumor images and quantified tumor burden using a newly developed semi-automated pipeline. Data from advanced melanoma patients receiving ICIs at the Lausanne University Hospital were then analysed using a tumor growth inhibition (TGI) model, aiming at evaluating the time course of tumor burden and quantifying the variability observed in a RW setting.
Methods:
We developed a semi-automated analysis workflow for collection and curation of RWD. Data aggregation and retrospective analysis for this study have been approved by the local Ethics Committee for patients who did not refuse general informed consent. Clinical data, treatment information and model covariates were extracted from electronic health records and semantically annotated to support cohort selection. Then, routine clinical imaging data (positron emission tomography-computed tomography: PET-CT and CT) were retrieved for the selected patients. Segmentation of the PET-CT scans relied on the PET-Assisted Reporting System prototype software (Siemens) [4, 5]. Data were further consolidated exploiting longitudinal imaging information. Radiology experts manually segmented CT scans.
Total tumor burden was computed for each patient by summing the volumes of all individual malignant lesions at each imaging exam and assessed in a TGI model developed in Monolix 2020R1. An exponential tumor growth with a rate constant retrieved from literature was assumed [2]. Different hypotheses on treatment dynamics were compared.
Results:
A total of 311 tumor volume measurements were available from 91 melanoma patients. The majority of the patients (n=38) received a combination of ipilimumab and nivolumab, potentially followed by a nivolumab maintenance phase (n=18). A TGI model using a log-kill hypothesis with a treatment effect linearly dependent on drug dose provided the best description of the data. Drug effects were included in an additive form. Results indicated the estimation of a shared killing rate parameter for nivolumab and pembrolizumab while a dedicated killing rate was estimated for ipilimumab. Inferring tumor exposure through a kinetic-pharmacodynamic model [6] or estimating separate residual errors for each image modality (i.e., PET-CT and CT) failed to improve the model. Model parameters were precisely estimated (relative standard error RSE < 21%) except for the ipilimumab killing rate (RSE 54%). A higher killing rate was estimated for nivolumab and pembrolizumab (0.0075 day-1) compared to ipilimumab (0.0027 day-1). This is in line with the higher clinical efficacy reported for PD1 inhibitors compared to CTLA4 inhibitors [7, 8]. Results also revealed large inter-individual variabilities (coefficient of variation >100%) in baseline tumor burden and killing rate constants. This is in accordance with previously published studies reporting high variabilities in tumor dynamics across melanoma patients receiving ICIs in RCTs [2, 3].
Conclusions:
By applying a semi-automated analysis pipeline to RW routine clinical data, we developed a TGI model that successfully described tumor dynamics in advanced melanoma patients. Our workflow constitutes a powerful tool to facilitate population-based studies using RWD, easily applicable to other cancer subtypes and therapies. To our knowledge, this is the first TGI model describing tumor response in patients receiving ICIs in a RW setting. Ongoing work involves the identification of sources of variability (including mutation status and radiomics features) to inform dosing strategies.
References:
[1] Regan MM, Mantia CM, Werner L, Tarhini AA, Larkin J, Stephen Hodi F, et al. Treatment-free survival over extended follow-up of patients with advanced melanoma treated with immune checkpoint inhibitors in CheckMate 067. J Immunother Cancer. 2021;9(11).
[2] Chatterjee MS, Elassaiss-Schaap J, Lindauer A, Turner DC, Sostelly A, Freshwater T, et al. Population Pharmacokinetic/Pharmacodynamic Modeling of Tumor Size Dynamics in Pembrolizumab-Treated Advanced Melanoma. CPT: pharmacometrics & systems pharmacology. 2017;6(1):29-39.
[3] Feng Y, Wang X, Suryawanshi S, Bello A, Roy A. Linking Tumor Growth Dynamics to Survival in Ipilimumab-Treated Patients With Advanced Melanoma Using Mixture Tumor Growth Dynamic Modeling. CPT: pharmacometrics & systems pharmacology. 2019;8(11):825-34.
[4] Capobianco N, Meignan M, Cottereau AS, Vercellino L, Sibille L, Spottiswoode B, et al. Deep-Learning (18)F-FDG Uptake Classification Enables Total Metabolic Tumor Volume Estimation in Diffuse Large B-Cell Lymphoma. J Nucl Med. 2021;62(1):30-6.
[5] Sibille L, Seifert R, Avramovic N, Vehren T, Spottiswoode B, Zuehlsdorff S, et al. (18)F-FDG PET/CT Uptake Classification in Lymphoma and Lung Cancer by Using Deep Convolutional Neural Networks. Radiology. 2020;294(2):445-52.
[6] Jacqmin P, Snoeck E, van Schaick EA, Gieschke R, Pillai P, Steimer JL, et al. Modelling response time profiles in the absence of drug concentrations: definition and performance evaluation of the K-PD model. Journal of pharmacokinetics and pharmacodynamics. 2007;34(1):57-85.
[7] Larkin J, Chiarion-Sileni V, Gonzalez R, Grob JJ, Rutkowski P, Lao CD, et al. Five-Year Survival with Combined Nivolumab and Ipilimumab in Advanced Melanoma. N Engl J Med. 2019;381(16):1535-46.
[8] Robert C, Schachter J, Long GV, Arance A, Grob JJ, Mortier L, et al. Pembrolizumab versus Ipilimumab in Advanced Melanoma. N Engl J Med. 2015;372(26):2521-32.
