2070 - What Can 5,045 Peripheral Lymphocyte Count Measurements in 255 Patients Tell Us About Temporal Dynamics and Survival in Non-Small Cell Lung Cancer?

V. Kaveckyte^1,2, A. Bjerrum¹, M. Pøhl¹, S. Vaabengaard¹, G. Persson^3,4, J. Petersen^1,5, A. J. Hope², T. Tadic², I. R. Vogelius^1,6, and J. Seuntjens²; ¹Department of Oncology, Rigshospitalet Copenhagen University Hospital, Copenhagen, Denmark, ²Department of Radiation Oncology, Princess Margaret Cancer Centre, University of Toronto, Toronto, ON, Canada, ³Department of Oncology, Herlev-Gentofte Hospital Copenhagen University Hospital, Copenhagen, Denmark, ⁴Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark, ⁵Department of Computer Science, University of Copenhagen, Copenhagen, Denmark, ⁶Deptartment of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark

Purpose/Objective(s): Published data suggest that peripheral lymphocyte counts (PLC) are associated with survival in locally advanced non-small cell lung cancer (LA-NSCLC) treated with radical concurrent chemoradiotherapy (CCRT). However, the importance of the temporal dynamics of PLC during and after CCRT is not well understood. A mathematical model describing PLC dynamics could mitigate bias from the exact timing of PLC measurements in multi-institutional studies. Here, we study the dynamics using 5,045 PLC measurements in 255 consecutive patients treated for LA-NSCLC and investigate the association with overall survival (OS). Materials/

Methods: The cohort consisted of patients diagnosed and treated Jan 2018 – Oct 2023 with a follow-up until Jan 2024. The patients received =1 cycle of platinum-based chemotherapy and 60-66 Gy in 2 Gy/fraction (4 patients received 55 Gy in 2.75 Gy/fraction). PLC were extracted from electronic health records (EPIC). Baseline PLC (week t = 0) were within 2 weeks before CCRT. To estimate individual end-of-RT (EoRT) PLC, the data were fitted to an exponential decay model PLC(t) = Ae^-Rt (t < t_EoRT) where A and R are fit parameters. 16 patients with <3 treatment data points were excluded. Patient-specific post-treatment PLC data were fitted with PLC(t) = PLC(EoRT) + C(1- e^{-R’(t-t_EoRT)}) (t_EoRT = t = 40) where C and R’ are fit parameters. 13 patients with <3 post-treatment data points were excluded. PLC(40) values were used to evaluate patient-specific PLC recovery from PLC(EoRT) relative to the baseline. OS was modeled by Cox regression using the baseline PLC measurement and the modeled PLC(EoRT) as predictors (uni- and bi-variate).
Results: The median baseline PLC was 1.8·10⁹ cells/L (1.4 – 2.2 IQR), the predicted median PLC(EoRT) was 0.6·10⁹ cells/L (0.4 – 0.8 IQR). The PLC measurements during CCRT agreed with the exponential decay model. The recovery from PLC(EoRT) was classified into 1) none to 30%, 2) 30-50%, and 3) =50% with approximately even distribution of 164 patients. The post-treatment PLC dynamics were heterogeneous, and the exponential model fit the recovery phase less well; 62 patients were excluded from classification due to non-convergence. More work is needed to classify PLC recovery for inclusion in Cox regression. We observed 93 events in 238 observations in the OS analysis after excluding an outlier (PLC(EoRT) > 9·10⁹cells/L). Baseline PLC could not predict OS, but PLC(EoRT) did predict OS (p=0.04 in bi- and p=0.03 in uni-variate analyses). Based on flexible modeling with splines most survival detriment is seen below 0.5·10⁹ cells/L (G3+ lymphopenia).
Conclusion: Temporal PLC dynamics during CCRT can be modeled with exponential decay, and PLC(EoRT) is prognostic of OS. Most patients’ PLC recovers at least partially but the recovery dynamics are heterogeneous. It is to be seen if post-treatment PLC dynamics can improve OS prediction. The immediate post-treatment PLC dynamics is of further interest given a hypothesis that PLC may predict immunotherapy efficiency.