Quantification of PD-L1 Expression with F-BMS-986192 PET/CT in Patients with Advanced-Stage Non-Small Cell Lung Cancer.

The aim of this work was to quantify the uptake of F-BMS-986192, a programmed cell death ligand 1 (PD-L1) adnectin PET tracer, in patients with non-small cell lung cancer. To this end, plasma input kinetic modeling of dynamic tumor uptake data with online arterial blood sampling was performed. In addition, the accuracy of simplified uptake metrics such as SUV was investigated. Data from a study with F-BMS-986192 in patients with advanced-stage non-small cell lung cancer eligible for nivolumab treatment were used if a dynamic scan was available and lesions were present in the field of view of the dynamic scan. After injection of F-BMS-986192, a 60-min dynamic PET/CT scan was started, followed by a 30-min whole-body PET/CT scan. Continuous arterial and discrete arterial and venous blood sampling were performed to determine a plasma input function. Tumor time-activity curves were fitted by several plasma input kinetic models. Simplified uptake parameters included tumor-to-blood ratio as well as several SUV measures. Twenty-two tumors in 9 patients were analyzed. The arterial plasma input single-tissue reversible compartment model with fitted blood volume fraction seems to be the most preferred model as it best fitted 11 of 18 tumor time-activity curves. The distribution volume ( ) ranged from 0.4 to 4.8 mL⋅cm Similar values were obtained with an image-derived input function. From the simplified measures, SUV normalized for body weight at 50 and 67 min after injection correlated best with , with an of more than 0.9. A single-tissue reversible model can be used to quantify tumor uptake of the PD-L1 PET tracer F-BMS-986192. SUV at 60 min after injection, normalized for body weight, is an accurate simplified parameter for uptake assessment of baseline studies. To assess its predictive value for response evaluation during programmed cell death protein 1 or PD-L1 immune checkpoint inhibition, further validation of SUV against based on an image-derived input function is recommended.