MRI radiomics for predicting pathological complete response in breast cancer: a systematic review of current evidence and limitations.

In 30 seconds

This systematic review evaluated MRI radiomics models for predicting pathological complete response (pCR) in breast cancer, analyzing 35 studies. The pooled area under the curve (AUC) was 0.813, indicating favorable discrimination. However, most studies were retrospective and underpowered, with limited external validation and subgroup analysis.

Key findings

• Pooled AUC for pCR prediction was 0.813, indicating moderate discrimination.
• 86% of studies were statistically underpowered, limiting reliability of findings.
• Only 14% of studies assessed performance across clinically relevant subgroups.
• Most studies were retrospective and single-center, affecting generalizability.

Why it matters

Understanding the limitations of MRI radiomics in predicting pCR is crucial for clinicians considering these models for patient management. Improved methodological rigor is necessary to ensure reliable and equitable application in diverse patient populations.