Comparing mammograms with multiple priors yields better detection, fewer recalls

Radiologists who compare new screening-mammography findings with those from two or more of the patient’s prior exams can expect better interpretive performance than radiologists who compare the new findings with those from only one prior exam.

The payoff will come in the form of significantly more true-positive results and significantly fewer false positives.

Researchers at UCSF Medical Center in San Francisco demonstrated the difference when they conducted a retrospective analysis of more than 45,000 consecutive screening mammograms performed at their institution for more than 22,000 women.

Their study posted ahead of print July 6 in the American Journal of Roentgenology.

Jessica Hayward, MD, and colleagues divided 46,288 consecutive screening mammograms into three groupings.

The first comprised mammograms interpreted without comparison with any prior exams.

The second grouping comprised mammograms interpreted in comparison with one prior exam—a protocol with established evidence for reduced recall rates, the authors note, but little to support better cancer detection.

The third grouping consisted of mammograms interpreted in comparison with two or more prior exams.

Defining positive predictive value level 1 (PPV1) as the percentage of screening exams with positive findings that resulted in a cancer diagnosis within one year, and defining cancer detection rate (CDR) as the number of cancers detected per 1,000 screening mammography examinations, the team found:

• The recall rate for mammograms that were compared with two or more prior exams was 6.3 percent—a marked improvement over both those without comparison (16.6 percent) and those compared with one prior exam (7.8 percent).
• Statistically significant increases in the PPV1 and the CDR occurred for screening mammograms compared with multiple prior mammograms versus a single prior mammogram: The PPV1 for the group with a single prior mammogram was 0.056 (95 percent confidence interval), while the PPV1 for the group with multiple prior mammograms was 0.105 (95 percent confidence interval).
• After adjustments were made for age, the odds ratio of recall for the group with multiple prior examinations versus group with a single prior examination was 0.864 (95 percent confidence interval).

In their discussion, Hayward et al. state that their findings confirm what they have observed anecdotally in their practice, where comparing new screening mammograms with at least two prior exams is standard operating procedure.

“We have based this approach on our experience,” they write, “which has indicated that comparison with prior mammograms may either obviate the recall of borderline findings that have a stable appearance or facilitate recall for those cases that show progressive change.”

The authors further note that the appearance of normal breast tissue often changes from year to year due to differences in mammography modalities and patient positioning.

In light of this, “the likelihood of establishing stability in the appearance of tissue increases with the number of prior examinations used for comparison, because the chance of identifying a prior examination in which the positioning and technique that were used are very similar to those used in the current examination is increased,” they add.

Similarly, “the likelihood of identifying progressive change—even slow change—in a real lesion increases as the number of prior examinations used for comparison increases.”