The radiologists who read mammograms in the United States face the daunting task of finding 3 to 6 cancers in every 1,000 mammograms. But despite the challenge, a recent study reported that most radiologists exceed performance levels set in the early 1990s.

The study, which appeared in the October Radiology, used data from the NCI-sponsored Breast Cancer Surveillance Consortium (BCSC) to evaluate how well 807 radiologists around the country were performing different aspects of breast cancer screening.

"We now have a baseline of where physicians are practicing," says lead author Dr. Robert Rosenberg of the Radiology Department at the University of New Mexico in Albuquerque. "If you look at cancer outcomes, the radiologists are doing a good job."

The performance levels set back in the early 1990s were basically educated guesses about how well a radiologist could do. Today, thanks to the BCSC, a wealth of data exists, and researchers are using the information to improve breast cancer screening.

The findings by Dr. Rosenberg's team provide performance "benchmarks" that may help radiologists identify which areas of their practices could be improved, says Dr. Stephen Taplin of NCI's DCCPS.

"Until you know where you stand, you can't figure out what you need to do to improve your performance," adds Dr. Taplin, who oversees the BCSC program.

Dr. Rosenberg's team analyzed 2.5 million mammography screening exams done between 1996 and 2002 involving 1.1 million women. They compared screening exam results with each woman's clinical outcome in the 12 months following her initial exam.

After follow-up work, cancer was diagnosed in 4.8 per 1,000 women. When a radiologist advised that a biopsy be performed immediately, 34 percent of the results indicated cancer.

The 807 radiologists came from nearly 200 different facilities that provide information on screening mammography to the BCSC.

The consortium, launched in 1994, includes five research sites currently gathering information on mammography through partnerships with facilities in their geographic areas. The facilities range in type from traditional radiology practices and hospital-based services to mobile mammography vans and pathology laboratories.

More than 225 published studies have used data collected by the BCSC.

"You could not do this particular study without the BCSC," notes Dr. Taplin. "It's very complicated to standardize data collection and definitions."

The findings suggest the need for more research on the recall rate - the proportion of women called back by radiologists for more evaluation based on a finding in a mammogram.

The recall rate should be less than 10 percent, and it has been higher in the United States than in Europe. The study reports a wide range, with some radiologists recalling 4 to 5 percent of women while others recall 20 percent or more.

"The question is whether we can narrow the range down," says Dr. Rosenberg, noting that the challenge will be to do so while detecting cancer at the same level.

A second study in the October Radiology used BCSC data to explore another aspect of screening mammography: whether radiologists are making the appropriate recommendations for follow-up care after assessing a mammogram.

This question is related to performance benchmarks because if the assessments are not properly linked to recommendations, then the benchmarks will be inaccurate and so will the audits that radiologists do to compare themselves to the benchmarks.

To answer the question, Dr. Berta Geller of the University of Vermont and colleagues analyzed several years of data before and after the final rules of the Mammography Quality Standards Act (MQSA) went into effect in April 1999.

These rules dictate the terminology that radiologists should use when making recommendations based on a finding in a mammogram. Previous studies found that assessments do not always match recommendations; the new study reports progress.

For example, in 1996 only 51 percent of the screening mammograms appropriately recommended a short-interval follow-up (usually 6 months) for women with lesions classified as "probably benign." By 2001, the number was up to 76 percent.

"The surprise is how quickly people adopted the new terminology," says Dr. Taplin.

Dr. Geller attributes the improvement to several factors, including greater familiarity with the terminology among radiologists due to the American College of Radiology's Breast Imaging Reporting And Data Systems (BI-RADS) and to the MQSA regulations.

More radiologists also are using computer software that automatically suggests recommendations when findings are entered on the computer. "This is a teaching tool, and radiologists who have difficulty with the terminology may find it useful," says Dr. Geller.

"Because breast cancer affects so many women and because mammography is the best current screening method, we're always trying to find ways to improve mammography," she says.

By Edward R. Winstead