Accuracy of mammographic breast density analysis: results of formal operator training

Mammographic density is a major risk factor for breast cancer. Breast density is not routinely quantified for research studies because present methods are time intensive and manual, and require expert training. We investigated whether individuals with and without backgrounds in radiology or medicine can achieve sufficient accuracy when compared with an expert (gold standard) reader of mammographic breast density. Nine readers (three radiologists, two non-radiology physicians, and four nonphysicians) assessed breast density on 144 digitized films (60 contralateral films of breast cancer cases and 84 controls) on a computer workstation with custom software. Readings were compared with a radiologist with training in mammography and density reading that read the same films. A correlation of r = 0.9 or higher with the gold standard reading was met by three of three radiologists, one of two nonradiology physicians, and one of four nonphysicians. Intrareader reproducibility measured as the residual sum of mean errors averaged 10% mammographic density for all readers and 9% for readers with a correlation of 0.9 or higher with the gold standard. The odds ratios associated with breast cancer when films with mammographic breast density of 50% or greater are considered "dense" ranged from 3.1 to 3.9 or a 1.9-2.4-per-population-SD increase in percentage density. Although it is advantageous to have a radiological background when quantifying mammographic density, it is not a prerequisite. Applying strict validation criteria to qualify readers to quantify mammographic breast density for research studies will enhance the chance of accurately assessing breast density and discriminating women at high and low risk of breast cancer.