Validation of Long-Term Overall Survival Estimates in Non-Small-Cell Lung Cancer Using Registry Data From the Danish Cancer Registry and the Impact of Improvement of Survival Outcomes Over Time

Speaker(s)

Metsä R1, Torkilseng E2, Vo L3, Harris M4, Milev S5, Sun A4, Smedegaard S6, Ulvestad M7, Lucherini S8
1Bristol Myers Squibb, Helsinki, 18, Finland, 2Bristol-Myers Squibb, Lysaker, Norway, 3Bristol Myers Squibb, Princeton, NJ, USA, 4Evidera, Bethesda, MD, USA, 5Evidera, San Francisco, CA, USA, 6Bristol Myers Squibb, Copenhagen, Denmark, 7Bristol-Myers Squibb, Oslo, Norway, 8Bristol Myers Squibb, London, CHE, UK

Presentation Documents

OBJECTIVES: CheckMate-816 (NCT02998528) met the primary objectives of event-free survival and pathological complete response after three doses nivolumab+chemotherapy relative to chemotherapy as neo-adjuvant (before surgery) treatment in patients with resectable NSCLC stage IB-IIIA (TNM version 7). A clinically important overall survival (OS) improvement trend was also observed. This study aimed to validate the modelled neoadjuvant chemotherapy OS using publicly available Danish registry data.

METHODS: A three-state semi-Markov model with pre-progression, progressed, and death health states was used. Transition probabilities were informed by extrapolation of trial data. A cure assumption was implemented in the model after 5 years for patients who had not died or progressed. Modelled OS was compared with a weighted Kaplan-Meier OS curve reflecting the CheckMate-816 baseline stage distribution based on OS curves stratified by pathological stage (pTNM) with 18+ years of follow-up from the Danish Lung Cancer Registry. Historical 1,2- and 5-year OS rates from patients treated with adjuvant chemotherapy reflective of clinical practice before of the introduction of neo-adjuvant immunotherapy were also used.

RESULTS: The absolute difference in OS rates years 1,2 and 5 for modelled neoadjuvant chemotherapy versus adjuvant chemotherapy treated patients were 1-3% for recently diagnosed versus 5-13% for patients diagnosed from 2003. The mean difference in OS in years 0-18 between the modelled neoadjuvant chemotherapy and the lung cancer registry long-term weighted Kaplan-Meier curve with and without a cure assumption were 12.45 and 4.47 months respectively.

CONCLUSIONS: Publicly available registry data can be a powerful source of information to inform validation of cost-effectiveness models but requires consideration of the impact of improvement in outcomes over time. Cancer registries should be encouraged to harmonize reporting of OS outcomes in their annual reports in a way that enable cross-country comparisons as well as validation of long-term OS according to TNM and relevant clinical variables.

Code

RWD151

Topic

Clinical Outcomes, Economic Evaluation, Methodological & Statistical Research

Topic Subcategory

Clinical Outcomes Assessment, Cost-comparison, Effectiveness, Utility, Benefit Analysis, Trial-Based Economic Evaluation

Disease

Biologics & Biosimilars, Oncology, Surgery