Frequency and predictive value of a mammographic recommendation for short-interval follow-up

Background: A recommendation for short-interval follow-up of "probably benign finding" is associated with up to 11% of screening mammograms, but its predictive value for breast cancer is unclear. We examined the predictive values (i.e., the percentage of women with a diagnosis of breast cancer 2 years after a short-interval follow-up recommendation) and likelihood ratios (derived from the pretest and post-test odds of breast cancer in the Women's Health Initiative sample) for breast cancer that are associated with a recommendation for short-interval follow-up among postmenopausal women. Methods: We performed a longitudinal analysis of a prospective cohort of 68126 postmenopausal women (aged 50-79 years) who were participating in clinical trials as part of the Women's Health Initiative at 40 centers across the United States. Eligible participants had screening mammograms at baseline and at least 2 years of follow-up that included a repeat mammography. Outcomes measured were breast cancer events at 1 and 2 years after baseline and the results of subsequent mammograms. All P values were two-sided. Results: A total of 2927 (5%) of the 58408 eligible women had baseline mammograms that included recommendations for short-interval follow-up. The incidence of breast cancer for women with a short-interval follow-up recommendation was 1.0% at 2 years after the baseline mammogram compared with breast cancer incidences of 0.6% and 0.5% for women whose baseline mammograms were described as "benign" and "negative," respectively. Across the 40 participating centers, the prevalence of short-interval follow-up recommendations among baseline mammograms varied from 1.2% to 9.8% (P < .001), even when the analysis was adjusted for key variables in regression models. Centers reporting higher frequencies of such recommendations did not have lower positive predictive values for breast cancer than centers reporting lower frequencies. The likelihood ratio for breast cancer after a recommendation for short-interval follow-up on a subsequent mammogram was 2.20 (95% confidence interval = 1.65 to 2.86). Conclusion: Having a mammographic recommendation for short-interval follow-up was associated with a low positive predictive value for breast cancer among postmenopausal women during a 2-year follow-up. This result suggests that the current criteria for this recommendation-repeat mammography within 6 months-should be reconsidered.