High quality mammograms are often misread and misinterpreted even by very experienced radiologists, according to research by Yale University School of Medicine (JAMA, 26 May 1993).

Ten "seasoned" radiologists were each given an identical set of 150 good quality mammograms to read. They had major clinical disagreements over the results such as normal versus cancer in a staggering 33 per cent of patients. And they disagreed "radically" in their recommendations of routine mammogram vs biopsy in a quarter of patients.

Interestingly, the researchers found no correlation between the radiologists' accuracy and their number of years experience or the number of mammograms they read a year.

Of the 150 patients used in the study, 27 had been "definitively" diagnosed as having cancer. The most accurate radiologist identified 23 of these as being "suspicious for cancer", and none of the other four were given "normal" results.

However, this radiologist also had the highest score for false positives and "frequently gave readings of cancer in patients who did not have it."

On the other hand, the radiologist who had picked up fewest of the genuine cancer cases (10 out of the 27) "had a false positive rate of only [italics ours] 20 per cent".