Double robust estimation of optimal partially adaptive treatment strategies: An application to breast cancer...

Area of Science:

• Biostatistics
• Computational Biology
• Precision Medicine

Background:

• Precision medicine seeks to personalize patient treatment based on individual characteristics.
• Existing double robust methods like G-estimation and dynamic weighted ordinary least squares (dWOLS) for optimal adaptive treatment strategies require modeling all treatment-confounder interactions.
• This complexity can be impractical for real-world applications.

Purpose of the Study:

• To develop and evaluate novel estimators for partially adaptive treatment strategies.
• To demonstrate the double robustness of these new estimators.
• To apply these methods to identify optimal hormonal therapy strategies for breast cancer patients.

Main Methods:

• Building upon G-estimation and dWOLS, new estimators for partially adaptive strategies were developed.
• The double robustness of the proposed estimators was theoretically demonstrated.
• A simulation study was conducted to investigate the performance of the estimators.
• The methods were applied to real-world breast cancer patient data from the Centre des Maladies du Sein.

Main Results:

• The proposed estimators were shown to be double robust.
• The simulation study validated the performance of the new methods.
• A partially adaptive treatment strategy for hormonal therapy in breast cancer was estimated using real patient data.

Conclusions:

• Partially adaptive treatment strategies, which tailor treatments based on a subset of covariates, are valuable in precision medicine.
• The developed estimators offer a more practical approach to identifying optimal treatment strategies compared to existing methods.
• The findings provide a new tool for tailoring breast cancer hormonal therapy.