Development of a Joint Tumor Size-Overall Survival Modeling and Simulation Framework Supporting Oncology Development...

Area of Science:

• Oncology
• Biostatistics
• Pharmacometrics

Background:

• Tumor size-overall survival (TS-OS) models are crucial for oncology drug development.
• Predicting long-term overall survival (OS) using early tumor size (TS) data and patient factors aids decision-making.
• Existing models require refinement for diverse treatment modalities.

Purpose of the Study:

• To develop and validate a robust TS-OS framework for predicting OS in non-small cell lung cancer (NSCLC).
• To identify key predictors of OS and TS dynamics across various cancer treatments.
• To establish a treatment-independent modeling approach for accelerated drug development.

Main Methods:

• Developed a framework jointly modeling TS (bi-exponential Stein model) and OS (accelerated failure time log-normal model).
• Utilized a treatment-independent link function connecting TS and OS.
• Incorporated baseline patient factors and tumor characteristics as covariates.

Main Results:

• Tumor growth rate (k_g) was the most significant OS predictor, modulated by an Emax function.
• Time to tumor growth and baseline TS also informed OS predictions.
• Albumin, total protein, and neutrophil-to-lymphocyte ratio were significant OS predictors.
• Baseline covariates influencing the TS model included target lesions, PD-L1 expression, and lactate dehydrogenase levels.
• The model accurately described OS distributions for chemotherapies, immuno-oncology, and combination treatments.

Conclusions:

• The developed TS-OS framework effectively predicts OS across diverse NSCLC treatments.
• The framework's treatment-independent nature supports its application in future study design and evaluation.
• This modeling approach can accelerate oncology drug development by providing reliable survival predictions from early data.