Mammographers continue to look for ways to improve reading methods: Investigators from Michigan studied outcome differences between next-day and same-day mammogram readings, and reported a higher positive predictive value (PPV) with the latter.
Meanwhile, a group from Vermont used independent double-interpretations of screening mammograms to see if the practice improved cancer detection, and found a relative increase in the number of early-stage cancers they were able to detect.
"Online screening mammography is performed with the prompt analysis of the mammogram by the radiologist on site. The patient is provided with immediate results of her mammogram," said Dr. Chintana Paramagul in a presentation at the 2003 American Roentgen Ray Society meeting in San Diego. "The objective of our study was to determine the differences of offline and online screening mammography and to compare the size of the lesion, biopsy rate, and call-back incidence."
The breast imaging center at the University of Michigan Health Systems in Ann Arbor currently offers asymptomatic women the option of offline (next day) or online (same day) reading. The benefit of the latter to the patient is that any additional imaging can be done at that time.
The retrospective study identified 9,653 women who had offline results and 9,653 with online results from January 1999 to December 2000. The same group of radiologists read mammograms from both groups and used the same equipment. At her institution, there are 10 radiologists who specialize in breast imaging, but one-third of these physicians read mammograms full-time, Paramagul wrote in an email to AuntMinnie.com.
According to the results, for offline screening, the callback rate was 10.8%, less than the 13.5% callback rate for online reading, for a 25% difference. The biopsy rate among offline screening patients was 1.9% versus 1.2% for online, meeting the published gold-standard biopsy rate of 2%-3%. At 23%, the PPV of online screening exceeded the PPV of offline screening (15%), again falling in the range of the 25%-40% gold standard, Paramagul said.
The cancer detection rate was 0.23% for offline reading and 0.16% for online. Fifty-eight percent of the cancers found offline were invasive, while 53.8% of those found online were invasive. Reading offline, the mammographers detected 3.8 cancers per 1,000, compared to 1.9 per 1,000.
For offline reading, 24% of the cancers detected were ductal carcinoma in situ, compared with 43% found online. The median size of the invasive cancer was 10 mm on the offline reads and 11 mm online. Finally, axillary lymph node involvement was found in 28% of the offline mammograms and in 27% of the ones read online.
Paramagul concluded that the number of cancers and the number of stage 0 and 1 tumors detected in this study surpassed that of national goals. However, she noted that because the women were allowed to choose whether they had online or offline reading, the study was not randomized. One bias was that there might have been better compliance with interval screening in the online group.
Independent double reading
For this study, two radiologists independently double-interpreted 25,369 screening mammograms performed from November 1998 and April 2000 at Fletcher Allen Health Care, University of Vermont College of Medicine, in Burlington.
"This study is unique because, to our knowledge, it is the first study in the United States to evaluate independent double interpretation of screening mammography in the clinical setting," wrote lead author Dr. Susan Harvey in the American Journal of Roentgenology. "In our study, screening was annual screening for women older than 40 years, with craniocaudal and mediolateral oblique views obtained of each breast," (AJR, May 2003, Vol. 180:5, pp. 1461-1467).
All mammograms were obtained using a Senographe 800T unit from GE Medical Systems (Waukesha, WI). The images were batch-hung on a dedicated mammography multiviewer with comparison films from 2-3 years earlier, when available.
On average, the first or the second radiologist interpreted 80 mammograms at one sitting. The first radiologist generally required 2.25 minutes per case to interpret films; the second took 0.5 seconds to 1 minute per case. In total, seven imaging specialists with 3-18 years worth of experience in mammography interpretation participated in the study.
"The first radiologist marked the findings on the mammogram if a recall for additional imaging was recommended. The second radiologist, or double-viewer, would then reinterpret the mammograms. The second reviewer looked at all the images and could add but not delete recalls," the authors explained.
Of the 25,369 screening mammograms, 14.2% of the women were recalled for additional imaging. The first reviewer recalled 12.7% of the cases while the second added an absolute recall rate of 1.5%. The first reviewer spotted 93.7% of the malignancies while the second reviewer caught an additional 6.3%.
"Our cancer detection rate with double interpretation was 5.6 per 1,000 screening mammograms and without double interpretation was 5.3 per 1,000 screening mammograms in the study population," the authors wrote. False-negative results were recorded in 1.5 per 1,000 screening mammograms.
With double interpretation, the sensitivity for cancer detection increased from 74.4% to 79.4%. A total of 672 biopsies were performed, with 38 generated by the second reader. Of those 38, nine were malignant.
"The relative rate of cancer detection was increased by 6.3% as a result of the second reviewers; and, with the exception of one cancer, all cancers detected by the second reviewer were stage 0 or I," the authors concluded. "These cancers were detected early, which can lead to lower mortality, lower cost, and the use of fewer resources."
However, the researchers also noted that because of the high recall rate reported in this study, they have modified their method of double interpretation from independent to consensus. They are currently collecting data from this change for further study.
By Shalmali PalAuntMinnie.com staff writer
May 21, 2003
Related Reading
Pain involved in mammography less stressful than waiting for results, April 15, 2003
Radiologists say training, not tests, will hone mammography skills, April 10, 2003
Patient notification letters: How to keep them hush-hush and HIPAA-compliant, April 4, 2003
Recently trained radiologists may interpret mammograms more accurately, February 27, 2003
Mammogram reading skills vary widely among community radiologists, September 18, 2002
Patients won't pay more for quick breast cancer screening results, September 6, 2001
Copyright © 2003 AuntMinnie.com
