Nataliya Kudryashova (1), Kirill Zhudenkov (1), Oleg Stepanov (1), Kirill Peskov (1, 2)
(1) M&S Decisions LLC, Moscow, Russia; (2) Computational Oncology Group, I.M. Sechenov First Moscow State Medical University, Moscow, Russia
Introduction: The socio-economic impact of cancer remains considerable and is poorly understood. It is associated with higher morbidity and mortality and should be addressed systematically [1]. Non-small cell lung cancer (NSCLC) is the most frequent cause of cancer-related death worldwide with ~1,800,000 deaths in 2020 [2]. In 2019, a total of ~50,000 lung cancer deaths produced an age-standardized mortality rate of 18.43 per 100,000 people in Russia [3]. Given the high significance of cancer for the healthcare system, application of model-based and meta-analysis approaches for the disease burden can provide health stakeholders with a tool to support decision making in resource allocation to maximize health benefit and decrease the cost of care and treatment [4].
Objectives: The primary objective of the study was to develop a methodology for patient survival analysis to perform a weighted indirect comparison of the disease burden in patients with solid cancers based on published epidemiological and clinical study data. The secondary objective was an application of the platform to assess the NSCLC burden in Russia and estimate the treatment efficacy according to the recently updated NSCLC local guidelines [5].
Methods: The platform development workflow was implemented as follows: published data collection, treatment scheme reconstruction, survival curve digitization, acquisition of patient cohorts’ characteristics, Markov chain model qualification and consequent survival analysis in the selected cohorts. Overall, over 200 publications were analyzed during NSCLC platform development. The whole patient cohort in the model was comprised of patients with early (stages I-IIIa [6,7]) and advanced (stages IIIb/IV) NSCLC, treated with a choice of therapy according to the local guides.
Published Kaplan-Meier curves were digitized using a modified approach proposed by Wei et al [8]. The survival data were further used to parameterize sub-models for each NSCLC stage as well as to describe survival of patients. A Markov chain model was qualified using Weibull transition probability functions to derive cohort survival for different patient subgroups (by treatment, cancer stage, therapy type, cancer histology, etc.)
Finally, for the whole patient cohort a set of treatment scenarios have been executed using monthly Markov cycles to inform patient survival as well as important cancer-related milestones (median survival, life-years gained (LYG), etc.) for up to 120 months.
Results: The compiled database for the model qualification included the data for early NSCLC stages (5 cohorts with respective disease stages for 3950 patients from epidemiology data) and advanced NSCLC (9 cohorts with respective treatments, 9 studies, 1585 patients from randomized clinical studies). Only studies with Caucasian patients with median age of 65 and ECOG 0-1 were included. The treatment scheme for NSCLC was reconstructed according to the local guides: patients could be treated with standard of care (chemotherapy), targeted (EGFR mutation present) or immunotherapy (TPS > 50%) [5].
For NSCLC patients a set of scenarios were simulated including specific biomarker/therapy and histology (squamous vs. non-squamous) parameters as well as variation of number of patients diagnosed at each stage. These scenarios influenced the cohort survival and provided a quantifiable output from advances in cancer treatment like enhancement of early cancer screening protocols or timely administration of specialized treatments. Thus, a comparison between the scenarios revealed that a complete implementation of recent updates in the local guidelines showed an overall gain in median population survival of 5 months for the patients with advanced NSCLC and provided additional 0.44 LYG at 60 months.
Conclusions: Here we presented a disease burden analysis platform to estimate population healthcare effects in solid tumor indications. The Markov chain model has the capacity to incorporate multiple stages of cancer as well as various treatment regimens to simulate patient survival in a mixed patient cohort reflecting actual epidemiological data on federal and regional levels. Further use of the platform can be extended to pharmacoeconomic and health outcome assessment of quality of life, utility, and treatment costs for patients with NSCLC.
References:
[1] Mehlis, K. et al. BMC Cancer 2020; 20, 529.
[2] Murray C. J.L. et al. N Engl J Med 2013; 369:448-457.
[3] https://gco.iarc.fr/today/data/factsheets/cancers/15-Lung-fact-sheet.pdf
[4] Kaprin, A. et al. 2019; Malignant neoplasms in Russia in 2018 (morbidity and mortality).
[5] Laktionov, K. et al. 2020; Malig. TUMOURS, Practical Guidelines RUSSCO 9, 32–48.
[6] Sui, X. et al. J. Thorac. Oncol. 2017; 12, 1679–1686.
[7] Yun, J.K. et al. J. Thorac. Dis. 2019; 11, 2955–2964.
[8] Wei Y & Royston P. Stata J. 2017;17(4):786-802.
Reference: PAGE 29 (2021) Abstr 9741 [www.page-meeting.org/?abstract=9741]
Poster: Drug/Disease Modelling - Oncology