Prediction of Tumor Control in ⁹⁰Y Radioembolization by Logit Models with PET/CT-Based Dose Metrics

Area of Science:

• Nuclear Medicine and Medical Imaging
• Radiation Oncology
• Medical Physics

Background:

• Radioembolization with ⁹⁰Y is a treatment for liver cancer.
• Accurate dosimetry is crucial for optimizing treatment outcomes and predicting tumor control.
• ⁹⁰Y PET/CT imaging allows for voxel-level absorbed dose (AD) calculations.

Purpose of the Study:

• To develop tumor control probability (TCP) models for radioembolization using ⁹⁰Y PET/CT-derived radiobiologic dose metrics.
• To assess the correlation between lesion absorbed dose and tumor shrinkage.
• To evaluate the relationship between non-tumoral liver dose and toxicity.

Main Methods:

• Patients with primary liver cancer or metastases underwent radioembolization with ⁹⁰Y.
• ⁹⁰Y PET/CT was used for voxel-level dosimetry to calculate AD and biological effective dose (BED) metrics.
• Logit regression models were employed to correlate lesion-level response (using mRECIST) with dose metrics.

Main Results:

• Tumor shrinkage showed a superior correlation with modified RECIST (mRECIST) compared to RECIST.
• The mean AD and BED metrics for 50% TCP were 292 Gy and 441 Gy, respectively.
• All evaluated dose metrics, including mean AD, were significantly associated with treatment response (AUC 0.87-0.90).

Conclusions:

• Logistic regression TCP models demonstrated a strong association between dose metrics and treatment response probability.
• Mean AD performed comparably to more complex radiobiologic dose metrics in predicting response.
• These findings underscore the importance of incorporating TCP into treatment planning for ⁹⁰Y radioembolization.