Evaluating double-reading mammography for long-term surveillance in breast cancer survivors: a retrospective exploratory analysis from a single center

Radiol Med. 2026 Feb;131(2):269-279. doi: 10.1007/s11547-025-02117-5. Epub 2025 Nov 6.

Abstract

Purpose: This exploratory analysis aims to assess the clinical feasibility of transitioning breast cancer (BC) survivors from intensive follow-up to a long-term surveillance strategy based on double-reading mammography within a population-based screening model.

Material and methods: Women who underwent BC surgery in 2013 and were followed for 10 years at a tertiary cancer center in Italy were retrospectively included. Recurrence rates and time to relapse were analyzed using the Kruskal-Wallis test and Chi-square test. Available follow-up mammograms were independently reviewed by three experienced radiologists, blinded to clinical information, using a positive/negative binary scale. Pathology served as the reference standard. Sensitivity, positive predictive value (PPV), and inter-reader agreement (Cohen's and Fleiss' κ) were calculated. Simulated double reading was analyzed using McNemar's test. A p value < 0.05 was considered statistically significant.

Results: Among 505 women, 46 (9.1%) experienced BC recurrence: 19/271 (7.0%) in Luminal A, 6/84 (7.1%) in Luminal B, 3/28 (10.7%) in triple-negative (TN), 7/57 (12.3%) in HER2-positive (HER2 +) cancers, and 11/65 (16.9%) in ductal carcinoma in situ (DCIS), which showed the highest recurrence rate (p = 0.0024). Median time to relapse ranged from 4.7 years (TN) to 8.1 years (Luminal A), with no statistically significant difference among the subtypes (p = 0.190). Individual reader sensitivity ranged from 67.7 to 74.2%, increasing to 77.4-83.9% with double reading. PPVs ranged from 95.4 to 100% for single reading and from 96.0% to 96.2% for double reading. Differences in sensitivity and PPV between single and double reading were not statistically significant (p = 1.0). Five recurrences (16.1%) were mammographically occult and clinically detected. Inter-reader agreement was moderate to good (Cohen's κ = 0.38-0.69; Fleiss' κ = 0.57).

Conclusions: Double-reading mammography showed a higher detection rate of BC recurrences compared to single reading in this exploratory setting, suggesting its potential role as a long-term surveillance strategy for BC survivors. However, the occurrence of relapses outside the mammographic field of view warrants careful consideration.

Keywords: Breast cancer; Double reading; Mammography; Recurrence; Surveillance.

MeSH terms

  • Adult
  • Aged
  • Breast Neoplasms* / diagnostic imaging
  • Breast Neoplasms* / pathology
  • Breast Neoplasms* / surgery
  • Cancer Survivors*
  • Feasibility Studies
  • Female
  • Humans
  • Italy
  • Mammography* / methods
  • Middle Aged
  • Neoplasm Recurrence, Local* / diagnostic imaging
  • Predictive Value of Tests
  • Retrospective Studies
  • Sensitivity and Specificity