Risk prediction models for malignancy upgrade in high-risk breast lesions: a qualitative systematic review

Area of Science:

• Oncology
• Pathology
• Biostatistics

Background:

• Atypical breast lesions on core needle biopsies are high-risk findings with uncertain clinical management due to variable upgrade rates to malignancy.
• Existing risk prediction models for malignant upgrade show inconsistent performance and predictor selection.
• This systematic review focuses on evaluating current models for predicting upgrade in high-risk breast lesions for clinical applicability.

Purpose of the Study:

• To systematically evaluate existing risk prediction models for malignant upgrade in high-risk breast lesions.
• To assess the clinical applicability and limitations of current prediction models.
• To identify common predictors and performance metrics of these models.

Main Methods:

• A qualitative systematic review following PRISMA guidelines was conducted.
• Searches in MEDLINE, Embase, and Scopus identified studies developing risk prediction models for malignancy upgrade after atypia diagnosis.
• The Prediction model Risk of Bias Assessment Tool (PROBAST) was used for quality assessment.

Main Results:

• Seventeen studies met inclusion criteria, with sample sizes ranging from 20 to 525 and upgrade rates from 14.9% to 67.3%.
• Common predictors included lesion size, histology, and radiologic-pathologic concordance; however, model performance varied significantly (AUROC 0.514-0.909).
• Most models lacked external validation, had poor calibration assessment, and exhibited a high risk of bias.

Conclusions:

• Current risk prediction models for malignant upgrade in high-risk breast lesions exhibit significant variability and limitations for widespread clinical use.
• These models may supplement clinical judgment but require further external validation and improved calibration.
• Reliable guidance for management necessitates enhanced model development and validation processes.