Radiologists' Mammogram Accuracy Varies Widely
Those at busiest centers better able to detect cancers, study found
FRIDAY, June 10, 2005 (HealthDay News) -- Researchers have found a wide gap in breast cancer detection between doctors who read a lot of mammograms annually and those who do not.
Based on a study of doctor's performance in interpreting more than 1 million mammograms from three states and the city of San Francisco, the team found that doctors with at least 25 years of experience who read more than 2,500 mammograms annually and who do general screening are about 30 percent better at detecting cancerous tumors than are doctors who read less than 780 mammograms a year.
The more experienced doctors have similarly lower rates of false-positive findings, which save women from unnecessary and invasive biopsies.
"This study highlighted the extreme variability between physicians in reading mammograms, and any woman ought to know that the accuracy she can expect varies widely," said lead researcher Philip Chu, an epidemiologist and specialist in the radiology department at the University of California, San Francisco.
However, he stressed that "mammograms are still the best method we have for catching cancer early on and saving lives."
Chu is to present his findings Saturday at the Department of Defense's Breast Cancer Research Program meeting in Philadelphia.
For the study, Chu and his colleagues focused on three predictors of accuracy in breast cancer detection: technology, the age of patients and their breast density, and the doctor's ability to accurately read mammograms.
They found that doctors accurately detected approximately 77 percent of the cancers, with the detection rate for individual doctors ranging from 29 percent to 97 percent. The average rate of false-positive readings was 10 percent, with rates for individual physicians ranging from 1 percent to 29 percent.
Overall, radiologists with more than 25 years experience who specialize in mammography and look at between 2,500 and 4,000 mammograms annually appear to have the best detection rates, the researchers concluded.
Chu said the message for women is that, if possible, they should have their mammograms done at large facilities where experienced doctors do a lot of screening for cancer every year, but said there is a message for doctors, too.
"To some degree, I think a lot of doctors aren't aware of their own accuracy," he said.
"I certainly agree with the findings, although this isn't news for us radiologists," said Dr. Ulana Suprun, vice medical director of Medical Imaging of Manhattan, which has three doctors reading more than 20,000 mammograms annually.
Suprun believes there are not enough radiologists to meet the growing demand for mammograms, primarily because poor reimbursement and high malpractice costs are discouraging young doctors from entering the field. So, while the best place to get a mammogram is at a busy facility, the main problem for many women is getting access to any facility.
However, she added that "no matter what facility to which you have access or who the doctor is reading the mammography, it is better to have one than not to have one. Mammography saves lives."
In another study reported at the conference, Duke University researchers described a computer program that could reduce the number of unnecessary biopsies needed in breast cancer diagnosis.
Radiologists currently have a standardized method of identifying the characteristics of lesions that they see on mammograms. By taking these methods and creating a computer algorithm, Anna Bilska-Wolak, a research associate at the Duke Advanced Imaging Laboratories, was able to increase the specificity of these characteristics to accurately identify all the malignant tumors and certain types of benign tumors, before a biopsy was done.
The tool could spare some women unnecessary biopsies, the researchers said.
The team ran 670 cases were run through the computer, 36 percent of which were malignant tumors and 64 percent which were benign. The tool correctly identified all the cancerous tumors and 32 percent of the benign tumors. In another evaluation of 150 cases, the tool again identified all the malignant tumors and 26 percent of the benign lesions.
Bilska-Wolak said she and her colleagues are continuing to test the new tool in more medical centers to see if those results are replicated.
For more on breast cancer, head to the American Cancer Society.