Bayesian Counterfactual Machine Learning Individualizes Radiation Modality Selection to Mitigate Immunosuppression.

Area of Science:

• Oncology
• Radiation Oncology
• Immunology

Background:

• Lymphocytes are crucial for cancer immunity and surveillance.
• Radiation-induced lymphopenia (RIL) is a common side effect of chemoradiation therapy (CRT), impairing immunity and worsening outcomes.
• Proton beam therapy (PBT) is hypothesized to reduce RIL risk compared to intensity-modulated radiation therapy (IMRT).

Purpose of the Study:

• To identify patient profiles benefiting from specific radiation modalities to mitigate RIL.
• To inform personalized radiation modality selection for cancer patients undergoing CRT.
• To investigate the impact of radiation modality on lymphocyte counts using advanced statistical methods.

Main Methods:

• A novel Bayesian causal inferential technique was applied.
• A matched retrospective cohort of 510 esophageal cancer patients undergoing CRT was analyzed.
• Patient profiles were identified based on factors like BMI, age, baseline absolute lymphocyte count (ALC), and planning target volume.

Main Results:

• Five distinct patient profiles were identified, showing varied responses to PBT versus IMRT.
• Significant differences in ALC nadir between PBT and IMRT were observed in three patient subtypes.
• Older, normal-weight patients (>69 years) had a two-fold greater ALC reduction with IMRT compared to PBT. Overweight/obese patients with lower baseline ALC (<1.6 k/µL) also experienced greater ALC reduction with IMRT.

Conclusions:

• Individualized radiation therapy selection is key to minimizing immunosuppression in high-risk cancer patients.
• Bayesian counterfactual modeling can identify complex patient profiles for clinical application.
• Personalized radiation strategies can improve outcomes by mitigating RIL during CRT.