May 1, 2012

Breast cancer screening: A trio of innovators shines light on the evidence women need

Few would argue the value of screening mammography in preventing deaths from breast cancer. But serious questions remain about how best to balance screening’s benefits with its potential harms, including false-positive test results, overdiagnosis leading to invasive treatment for non-life-threatening conditions, and repeated exposure to medical radiation.

The uncertainty has both the U.S. Preventive Services Task Force and the Institute of Medicine calling for better evidence about screening outcomes  and the myriad factors that affect them. Through groundbreaking research with the Breast Cancer Surveillance Consortium (BCSC), GHRI Senior Investigators Diana Miglioretti, PhD, and Diana S.M. Buist, PhD, MPH, and Assistant Investigator Rebecca Hubbard, PhD, provided that evidence in 2011.

The BCSC is a national network of seven mammography registries with links to tumor and/or pathology registries. Established in 1994, it is the nation’s largest longitudinal collection of breast imaging data—a one-of-a-kind  resource that lets researchers worldwide answer timely questions about benefits and risks of breast cancer screening.

One recent example: while annual mammograms have long been standard for women with a personal history of breast cancer, only scant evidence had supported the practice. But with the BCSC’s comprehensive data, Dr. Miglioretti—the Consortium’s co-principal investigator—worked with Dr. Buist and colleagues from the University of Sydney in Australia on the first-ever study to assess screening mammography outcomes rigorously for women who’d had breast cancer before. Examining 12 years of data from more than 110,000 mammograms, they found that yearly screening detected second breast  cancers at an early stage—results they published February 23, 2011 in the Journal  of the American Medical Association.

In April 2011’s Radiology, Dr. Buist and BCSC colleagues published evidence from the largest study of U.S. radiologists to date, showing that interpreting more screening mammograms helps radiologists do a better job  of determining which suspicious breast lesions are cancer. They also found that  reading diagnostic mammograms helped improve interpretive accuracy of  screening exams. The findings have important policy implications: The Food and Drug Administration currently requires that only 960 screening mammograms be read over two years. By comparison, European and Canadian requirements range from 2500 to 5000.

“Based on our data, increasing volume requirements and including a minimal requirement for diagnostic interpretation would likely help radiologists find more cancers early and reduce the number of false-positive findings,” explains Dr. Buist, who leads Group Health’s breast cancer registry.

False positives lead to $1.6 billion in health care costs each year—while causing stress and anxiety for millions of women. That’s why Dr. Hubbard and colleagues used BCSC data to examine false-positive results from  screening mammograms for 169,000 women. In the October 18, 2011 Annals of  Internal Medicine, they reported that, on average, more than half of women screened yearly for a decade had a false-positive result—and that having prior mammograms for comparison cut the odds of a false-positive in half. Screening every two years instead of annually lowered the false-positive rate by a third.  Starting screening at age 50 instead of age 40 also lowered the lifetime risk of false positives.

“The BCSC’s goal is to take screening recommendations from one-size-fits-all to personalized,” Dr. Miglioretti explains. Thanks to the Consortium’s new $20 million Program Project grant from the National Cancer Institute, that goal is within reach.

GHRI’s statistical leadership brings data to life  

Integral  to the BCSC’s success is its ability to collect, store, and analyze data across diverse registries. By 2011’s close, the BCSC database included 9.5 million mammograms, 2.3 million women, and 114,000 breast cancer cases. Those quantities keep rising, as do the numbers of grant applications and publications supported by BCSC data—more than 70 and 400, respectively, to date. Most important, BCSC data play an essential role in breast cancer screening policy decisions and clinical recommendations. How do data achieve such influential reach?

“Our Statistical Coordinating Center—or SCC—pools the data from participating registries,” explains Dr. Miglioretti, a nationally recognized biostatistician and the SCC’s principal investigator. “We establish and evaluate data collection, help individual sites and outside researchers with analysis, and develop new statistical techniques.” Recognizing the SCC’s broad influence, NCI awarded the team a five-year, $7.5 million contract in 2011 to make BCSC data more widely available.

Dr. Miglioretti and her team are not alone in bringing GHRI’s data savvy to national consortia. Associate Investigator Jennifer Nelson, PhD, serves as co-lead of the methods core for the Food & Drug Administration’s Mini-Sentinel, a collaboration of 25 organizations building a national electronic medical product safety monitoring system. Many GHRI investigators, biostatisticians, and programmers have also made fundamental contributions to the HMO Research Network’s virtual data warehouse—a  standardized federated data system that supports and streamlines multi-center  research projects.


Jennifer C. Nelson, PhD

Director of Biostatistics; Senior Investigator
Kaiser Permanente Washington Health Research Institute