Breast density scales: the metric matters.

Area of Science:

• Radiology and Medical Imaging
• Biostatistics
• Oncology

Background:

• Percent mammographic density (PMD) is often categorized using various scales.
• This categorization may lead to a loss of valuable information.

Purpose of the Study:

• To examine the information loss associated with using categorical density scales for PMD.
• To evaluate the impact of PMD categorization on breast cancer risk models.

Main Methods:

• Assessed baseline PMD at 1% precision for 2,374 females.
• Created 21, 4, and 2-category density scales from continuous PMD data.
• Evaluated categorization effects using R-squared and root mean square error.
• Compared cancer risk model performance using categorical vs. continuous PMD via ROC curves.

Main Results:

• R-squared decreased significantly from 1.00 (1% PMD) to 0.56 (2-category scale).
• Root mean square error increased from 0.00 (1% PMD) to 10.83 (2-category scale).
• Area under the ROC curve for risk models decreased with increasing categorization, with a 2-category scale showing no discriminatory power (0.50).

Conclusions:

• Categorizing PMD into density scales results in substantial information loss.
• Continuous PMD measures are superior to categorical scales for breast cancer risk modeling.
• Simplified PMD categories, especially binary splits, severely impair the predictive accuracy of risk models.