OBJECTIVES

: Immunotherapy is known to provide a substantial survival benefit to a select group of non-small cell lung cancer (NSCLC) patients, therefore, early identification of progression provides a more beneficial treatment to these patients. Serum tumor markers can be used to monitor the response to treatment, however no clear guidance is available how to use and interpret their longitudinal results. Our study aims to evaluate and compare longitudinal biomarker analysis methods, in early detection of progressive disease in immunotherapy treated NSCLC patients.

METHODS

: A cohort of 434 NSCLC patients treated with nivolumab or pembrolizumab provided bi-weekly analysis of CYFRA, CEA, CA125, NSE, and SCC as part of regular care. Disease progression was determined using RESIST criteria and clinical assessment. Six logical and statistical methods based on longitudinal biomarker analysis and interpretation were evaluated on their ability to identify patients presenting with progressive disease at six months, 1) two consecutive increments, using baseline or the first of two compared results as a reference, 2) biomarker doubling time, 3) the slope between two consecutive results, 4) the absolute change between the baseline measurement and the measurement at week-6, 5) a cox proportional hazards model, 6) a landmark model.

RESULTS

: The sensitivity of each method was determined at a constant specificity of 95% to ensure false-positive rates remained low. On average, the sensitivity ranged from 0% for CEA in the cox model to 26% for CYFRA when looking at the change between baseline and week-6.

CONCLUSIONS

: Our research demonstrates that different models to interpret longitudinal serum biomarker measurements for NSCLC have different characteristics and resulted in different diagnostic performances. Models with the best diagnostic characteristics might be valuable monitor immunotherapy response in NSCLC patients.