Integration of early circulating tumour DNA (ctDNA) changes into tumour growth inhibition modelling.
Jinju Guk1, German Leparc1, Raphael Hesse2, Neetika Nath1, Eva Germovsek1, Lisa Amann1, Annika-Maria Mueller2
[1] TMCP Data Sciences, Boehringer Ingelheim Pharma GmbH & Co. KG [2] TMCP Translational Medicine & Biomarker Platforms, Boehringer Ingelheim Pharma GmbH & Co. KG
Introduction: Circulating tumour DNA (ctDNA) has been considered a surrogate marker to predict clinical response in several types of cancer and early clearance of ctDNA has been studied in this context [1-3]. However, there is no standard quantity and summary metrics of ctDNA as surrogate of clinical responses for different cancer indications or treatments. Moreover, considering specific mutation types into pharmacometric modelling might be challenging due to high dimensionality.
Objectives: In this study, we aimed to explore the implementation of early changes of ctDNA into pharmacometric tumour growth inhibition modelling using two summary indices: maximal allele frequency [4] and Shannon index [5,6].
Methods: Data from 35 patients with advanced solid tumours treated in a phase 1 dose escalation trial including 132 longitudinal tumour measurements (as the sum of the longest diameter of target lesions (SLD)) and 185 ctDNA samples were analyzed. Variant allele frequency (VAF) was determined using the NGS-based Roche Avenio expanded panel (77 genes). According to a recent study, the maximal VAF was calculated using all somatic tumour-derived variants for each unique patient sample regardless of the presence at baseline [4]. In addition, Shannon index was calculated using detected VAF to reflect heterogeneity of tumour cells [5,6]. Using the two calculated indices, two metrics – the area under the curve of absolute (AUCA) and fold changes from the baseline (AUCF) – were calculated. Only ctDNA measurements up to cycle 1 were used to represent early changes before the first imaging scan scheduled at the end of cycle 1. As an exploratory step, the proxy for tumour change rate (kproxy) was calculated as the minimal value of SLD change rate from the baseline of each patient and attempted to correlate to AUCA and AUCF. The model by Claret at al. [7] was used as a base tumour growth inhibition model and AUCA and AUCF of both VAF and Shannon index were tested using an exponential function. R 3.5.2 was used to calculate AUCA and AUCF, and NONMEM 7.4.3 was applied for parameter estimation.
Results: In the data exploration stage, the AUCF of maximal allele frequency was significantly (p < 0.01) correlated to the kproxy. In addition, a correlation with the Shannon index was observed, although not significant. The AUCA of both indices was not correlated to kproxy . In the model implementation stage, the longitudinal tumour growth profile was well characterized with high inter-individual variability in tumour growth rate (coefficient of variation (CV) (%): 80.5 %) and tumour killing rate (CV (%): 88.8 %). No inter-individual variability was considered in the resistance parameter. In the covariate building step, as expected, either maximal variant allele frequency or Shannon index turned out to be significant covariates on the killing rate. The covariate coefficient was -8.32 (relative standard error (r.s.e.), 25.5 %) for maximal VAF (p-value < 0.001) and -0.0595 (r.s.e. 36.3 %) for Shannon index (p-value < 0.01)) as an exponential relationship. Approximately, 27 % or 9 % of CV in the killing rate was reduced by either maximal VAF or Shannon index changes, respectively. These results indicate that patients having an early decrease in ctDNA will likely have better tumour responses, which is consistent to observations in previous studies [1-3].
Conclusion: As a top-down approach, transformed one-dimensional index of ctDNA data could be integrated in longitudinal tumour growth inhibition models to potentially increase predictivity of highly variable tumour response to the treatment.
Disclosure: The authors meet criteria for authorship as recommended by the International Committee of Medical Journal Editors (ICMJE), and did not receive payment related to the development of the abstract. BI was given the opportunity to review the manuscript for medical and scientific accuracy as well as intellectual property considerations. The study was supported and funded by BI.
[1] Pessoa LS et al.,ctDNA as a cancer biomarker: A broad overview. Crit Rev Oncol Hematol. 2020;155:103109. doi:10.1016/j.critrevonc.2020.103109
[2] Song Y et al. Circulating tumor DNA clearance predicts prognosis across treatment regimen in a large real-world longitudinally monitored advanced non-small cell lung cancer cohort. Transl Lung Cancer Res. 2020;9(2):269-279. doi:10.21037/tlcr.2020.03.17
[3] Bratman SV et al. Personalized circulating tumor DNA analysis as a predictive biomarker in solid tumor patients treated with pembrolizumab. Nat Cancer. 2020;1(9):873-881. doi:10.1038/s43018-020-0096-5
[4] Vega DM et al. Changes in Circulating Tumor DNA Reflect Clinical Benefit Across Multiple Studies of Patients With Non-Small-Cell Lung Cancer Treated With Immune Checkpoint Inhibitors. JCO Precis Oncol. 2022;6:e2100372. doi:10.1200/PO.21.00372
[5] Yaung C et al., Ecological diversity indices as measurements of tumor heterogeneity correlates with clinical outcomes in late state small cell lung cancer (SCLC). Annals of Oncology 29(suppl_8). 2018;Oct. doi:10.1093/annonc/mdy269.168
[6] Shannon CE and Weaver W. The mathematical theory of communication. University of Illinois Press. (1949).
[7] Claret L et al. Model-based prediction of phase III overall survival in colorectal cancer on the basis of phase II tumor dynamics. J Clin Oncol. 2009;27(25):4103-4108. doi:10.1200/JCO.2008.21.0807
