Statistical methods applied to omics data: predicting response to neoadjuvant therapy in breast cancer

Area of Science:

• Oncology
• Genomics
• Biostatistics

Background:

• Omics technologies are vital in oncology clinical trials for understanding molecular mechanisms and identifying therapeutic targets.
• High-dimensional omics data often pose challenges for standard statistical methods.
• Accurate modeling of omics data is essential for identifying significant genes.

Purpose of the Study:

• To review statistical methods used for gene identification in oncology clinical trials.
• To compare penalized methods with newer causal inference methods for omics data analysis.
• To illustrate these statistical approaches in a neoadjuvant breast cancer trial.

Main Methods:

• Systematic review of statistical methods in 13 recent oncology publications.
• Focus on penalized methods and causal inference techniques.
• Application of selected methods to a nonrandomized neoadjuvant phase II trial in breast cancer patients.

Main Results:

• Most reviewed studies had small sample sizes and were non-randomized.
• Penalized methods are commonly used for gene identification.
• In the illustrated trial, causal inference methods did not outperform penalized methods.

Conclusions:

• Oncology clinical trials often feature small sample sizes and a wide array of statistical methods.
• Penalized methods are a robust choice for analyzing high-dimensional omics data.
• Causal inference methods may not offer superior performance over penalized methods in certain settings.