This monocentric study (Breast Unit,
Trieste-Italy) included 150 consecutive recalled women from the DM screening program between January and May 2018.
The cases that presented microcalcifications as mammographic pattern were excluded.
All recalled women were evaluated with DBT as second level examination.
A database has been constructed containing the regional code (CRA) for each woman,
as well as the age at the time of the examination,
the date of execution of the mammograms,
the breast density stated on the report (according to BI-RADS®36),
the reason for the recall,
the side and site of the lesion,
the BI-RADS® evaluation category,
the date of execution of the second level examination,
the type of investigation carried out specifying projections of tomosynthesis and the eventual magnification or compressions in the course of DBT or DM,
the BI-RADS® category after second level,
the result of needle aspiration (FNAC) or vacuum assisted biopsy (VAB) or needle biopsy (TRU-CUT),
the outcome of MRI if performed and the final diagnosis after the second level exams.
Screening mammograms for the regional program were obtained with the Planmed Nuance ™ Classic indirect digital mammography system.
The tomosynthesis in patients recalled from screening was performed with the Selenia® Dimensions® Hologic® mammography system with an integrated tomosynthesis module,
at the Breast Imaging section of Department of Radiology of Cattinara Hospital in Trieste.
The retrospective review of the images involved two radiologists with different experiences: radiologist A,
with twenty-five years of experience in breast imaging (also reader of the regional screening program for twelve years),
and radiologist B,
with ten years of experience in breast imaging.
Independently each radiologist performed a double-blind revision of all DBT acquired in the Screening Assessment Clinic.
A folder number was created to refer to the patient,
in order to hide any information about the woman and her medical history.
To allow comparison between one and two view DBT,
reading images were made for each patient in two steps: first in the mediolateral or mediolateral-oblique projection (ML or MLO) and then in craniocaudal (CC).
The characteristics of the women subjected to deepening through DBT have been summarized by means of descriptive analysis.
Continuous variables have been expressed as media and standard deviation or median and range (minimum-maximum).
The McNemar test for coupled data and Cohen's Kappa coefficient were calculated to evaluate the agreement between the two radiologists both in one and two view DBT.
Evaluation of DBT as a diagnostic test in identifying malignant or benign lesions correctly,
considering as gold standard the cytology or histology in the case of carcinomas and cytology or histology or other tests of in-depth examination (mammography,
ultrasound) for benign lesions,
was carried out calculating the diagnostic performance indicators such as: accuracy,
sensitivity,
specificity,
positive predictive value and negative predictive value with their intervals 95% confidence.
The proportions test was used to compare the diagnostic performance indicators between the two radiologists or between the two different types of reading (one or two view).
The analysis of inferential statistics was deepened using the ROC curve (Receiver Operating Characteristics) by calculating the area under the curve (AUC).
The analysis of the ROC curve was used to evaluate the diagnostic performance of DBT according to the type of reading (one view vs.
two view),
of the radiologist (radiologist A vs radiologist B),
density (A-B vs C-D) and mammographic pattern type (opacity vs asymmetry / distortion).
The differences between the underlying areas of the ROC curves were evaluated by the DeLong test.
The analysis was performed using software R version 3.5.0 (The R Foundation for Statistical Computing) and STATA 14.2 (StataCorp,
College Station,
Texas).
All values of the p-value (p) refer to statistical tests with two queues and if not otherwise specified,
the differences were considered statistically significant for values of p <0.05.