The exclusive and systematic use of the DBT in breast cancer diagnostics improved the diagnosis accuracy, in terms of sensitivity and specificity, without significantly increasing the amount of radiation that patients undergo. Specifically, DBT has proved to be extremely sensitive as it is able to highlight small masses and distortions better than 2D FFDM, and particularly allows better detection of benign lesions from images without the need for biopsies.

Several state-of-the-art studies have already compared the FFDM with the DBT in terms of sensitivity and specificity, as well as in terms of Benign Biopsy Rate and radiologists’ performances. In this study, conversely, we compared the accuracy of the two diagnostic techniques in terms of malignant (B5) Biopsy Rate and PPV-3, particularly focusing on data related to the VABB, which is the biopsy technique generally used for in-depth analysis of mammogram detected lesions. Thus, after collecting 69,384 mammograms and 7894 biopsies performed from 2012 to 2021 on female patients afferent to the Breast Unit of the Istituto Tumori “Giovanni Paolo II” of Bari, we estimated three linear regression models which describe the overall Biopsy Rate, CBs Biopsy Rate, and VABBs Biopsy Rate variations, over the 10 year period of activity. Even though we expected the stereotactic biopsy results to be the most significant, we intentionally also reported data related to the Core Biopsies, comparing each one over the 10 year period and considering the integration of CBs which were performed on the same period.

Finally, we compared the performances of three breast radiologists at our institute to verify their individual performances in detecting neoplasms.

As a result, the first linear regression model highlighted an overall reduction of the total number of biopsies carried out during the 10 years and proved to be statistically significant with a p-value equal to 0.04. This result was also confirmed by the estimated linear regression model for analyzing the VABB Biopsy Rate variations, on the contrary, the experimental results show no significant reduction in the Core Biopsy Rate. Although the number of cases referring to the VABB procedure is significantly lower than that of the CBs, we feel that the result which emerged might well explain the significant reduction observed in the overall sample.

As a matter of fact, there was a relevant reduction in VABBs carried out during this period, particularly after the systematic introduction of DBT in 2017 when the VABB Biopsy Rate decreased from 2.7 to 1.6%: and the estimated model resulted statistically significant with a p-value of less than 0.005. Besides, we demonstrated that the decrease in the number of VABBs was not as the consequence of a reduction in B5 tumors diagnosed, by means of a further linear regression model which proved not to be statistically significant (p-value equal to 0.69).

Our analysis demonstrates that the systematic introduction of DBT in our institute reduced the overall number of biopsies required after the first mammographic screening without compromising the diagnostic accuracy. On the basis of this, patients avoid further unnecessary examinations and costs (related to breast cancer diagnosis) are reduced. In the light of these encouraging results, in a future study, we will go on to analyze data collected across a multicenter study to confirm the significant accuracy of DBT compared to 2D FFDM in terms of Malignant Biopsy Rate and PPV-3.

It should be emphasized that our study refers to the experience of “Giovanni Paolo II” Cancer Institute alone, which has been the regional oncological reference hub for breast pathology for the last decade. Nevertheless, our findings may not be representative of other realities.

Our results supported the literature, showing that DBT simultaneously improves breast cancer detection by reducing false positive recalls with fewer biopsies performed.

A limitation could be the radiation dose which, on average, is about 30% higher in the DBT plus SM (Synthesized Mammography) protocol compared to the FFDM, data which, however, are considered acceptable and in line with the European guidelines for quality assurance in mammography screening. Additionally, the systematic use of tomosynthesis could mean a longer reading time and decreased productivity. However, in our experience, an increase in mammographic performance was observed over time for the same number of operators, with the sole exception of the first part of the COVID period: some retrospective studies in the literature have also shown that radiologists using AI tools for the simultaneous reading of DBT or a combination of several techniques could reduce the reading time by maintaining a level that is not lower or obtaining better performances in breast cancer diagnosis [2, 6, 9,10,11, 36] as already documented on other neoplasms [4, 20, 21, 38, 43]. A further limitation of DBT is overdiagnosis, defined as the detection of multiple low-grade lesions and small tumors, which carries economic implications that need to be considered.

Future studies to explore breast cancer growth and access to long-term clinical follow-up data are needed to fully understand the complex question of identifying and treating small and slow-growing breast cancers associated with DBT screening [23].