Tissue Validation of Mammographic Breast Arterial Calcification in Mastectomy Specimens

This is a single-centre, prospective, cross-sectional diagnostic accuracy study. There is no randomisation, no control group and no intervention.

Rationale Artificial-intelligence-based quantification of breast arterial calcification (BAC) in a cohort of 123,762 women has shown a dose-response relationship with major adverse cardiovascular events and incremental prognostic value over conventional risk scores. Systematic reviews support this association. Two questions remain unresolved. First, BAC is thought to be located in the tunica media rather than the tunica intima, making it medial arterial calcification rather than atherosclerosis, but direct histopathological verification in breast tissue is very limited. Second, and more critically, the field has no external reference standard: reported sensitivity values measure agreement with a human reader, not with tissue.

Specimen Handling The operation is not modified for the study. The specimen is oriented in theatre with two sutures (single suture superior, double suture lateral). In addition to diagnostic sampling, six research blocks are taken from macroscopically tumour-free areas: two from the upper outer quadrant, one from the upper inner quadrant, one from the lower quadrants, one from deep tissue adjacent to the prepectoral fascia and one from the central or retroareolar region. If there is any doubt about the adequacy of tissue for diagnosis, no additional block is taken; correct diagnosis always takes precedence.

Histopathological Assessment A stepwise staining algorithm is used: routine haematoxylin and eosin on all six blocks; von Kossa on blocks containing arteries of assessable calibre; elastic van Gieson on cases with calcification, to determine the layer involved. Presence, location (medial or intimal), circumferential extent and a 0-3 ordinal severity score are recorded, together with morphometric measurements made in free software. Intimal lesions are graded 0-4 using predefined morphological criteria, with grade 2 or above classified as atherosclerotic change. No additional immunohistochemistry is performed.

Mammographic Assessment The existing preoperative digital mammogram is read primarily in the mediolateral oblique projection. Presence of BAC, a 0-3 ordinal severity score and an estimated calcification area in square millimetres are recorded.

Coronary Artery Calcification Ordinal scoring is performed on staging positron emission tomography-computed tomography (PET-CT) or thoracic computed tomography (CT) images already acquired. The reader is a radiologist or nuclear medicine physician who is not among the mammography readers of this study. Cases without thoracic imaging, and those with previous coronary revascularisation, are excluded from this sub-analysis but remain in the primary analysis.

Blinding The pathologist is blind to mammographic and coronary findings. The mammography reader is blind to histopathology and coronary findings. The coronary calcification reader is blind to histopathology, mammography and the patient's clinical cardiovascular history. Twenty per cent of cases are re-read by independent second readers to measure inter-observer agreement.

Statistics With histopathology as the reference standard, sensitivity, specificity and predictive values of mammography are estimated with 95 per cent confidence intervals by the Wilson method. Because this is a preliminary study, the primary inference is drawn from the upper limit of the confidence interval rather than from the point estimate. Reporting follows the Standards for Reporting of Diagnostic Accuracy Studies (STARD) guideline.