The clinical implementation of DBT in both diagnostic and screening breast imaging has improved the cancer detection rate, and in particular the detection of ADs [27, 28], due to the reduction in the superimposition of overlapping breast tissues, in comparison to FFDM.
On the contrary, the benefits of a higher CDR have led to drawbacks in terms of increased screening recall rates and need for invasive assessment in recalls from screening [7, 8]; in particular the “Verona Trial” by Caumo et al. reported a CDR of 9.30 per 1000 screening examinations of DBT plus synthetic 2D vs. 5.41 per 10000 of FFDM alone, with comparable recall rates, but increased reading time and higher recall rates with invasive assessment [7].
Lång et al., illustrated similar results in the ‘Malmo Trial’, concluding that one-view DBT may be feasible as a stand-alone technique for breast cancer screening since DBT has proven to be superior in terms of detection rate and equal in term of PPV if compared with FFDM but with increased recall rate, even if still low, 3.8 for DBT vs. 2.6 for FFDM (p < 0.0001) [8].
This disadvantage is particularly impactful in cases where the abnormality detected is an AD, with the result of a lower PPV for malignancy, in particular in a meta-analysis of Choudhery et al., the PPV of ADs seen only on DBT images is 34.6% [29]. Many authors focused their efforts on identifying some imaging features that could help the prediction of malignancy of a newly-detected AD [30], to reduce the need for invasive procedures and further assessments.
Breast functional imaging is recently making his way as an adjunct tool to differentiate between benign and malignant ADs and to potentially avoid unnecessary biopsies. Both breast MRI and CEM are two functional imaging techniques that highlight lesions vascularity by means of contrast medium administration [16].
CEM has proven to have an optimal performance in terms of sensitivity and specificity, superior to that of FFDM and equivalent to that of MRI, with a slightly superior specificity vs. MRI [31, 32]. In our series, CEM showed a very high NPV of 99.37%, suggesting that BI-RADS 3 ADs that do not show enhancement may be safely sent to follow-up with a high level of confidence, avoiding unnecessary biopsies. We reported just one FN case in a 50-year-old intermediate risk patient: the AD in this FN case corresponded to a 5 mm invasive lobular carcinoma in a CEM examination that was reported as negative because of marked BPE that we believe could have masked the enhancement of the index lesion.
Some previous studies analyzed the performance of breast MRI in the assessment of suspicious or equivocal ADs with optimal NPVs [33, 34]; in particular a meta-analysis by Ferre et al., confirmed a NPV of 98.3–100% concluding that with a negative MRI the biopsy can be avoided in favor of a short-interval follow-up [15]. Another study by Mei et al., investigated the diagnostic value of MRI for architectural distortions categorized as BI-RADS 3–4 by FFDM and demonstrated that MRI could improve the diagnostic efficacy of FFDM [14].
To our knowledge, up to date few studies have analyzed the role of CEM in the assessment of BI-RADS 3 AD. In the study of Goh et al., the authors analysed CEM examinations of 94 ADs detected by FFDM and DBT and reported a sensitivity, specificity, PPV and NPV in predicting malignancy of respectively 100%, 42.6%, 48.5% and 100%, concluding that the absence of enhancement was highly indicative of no underlying malignancy [33]; a second study from Patel et al., included 49 ADs detected by DBT-only and reported a sensitivity, specificity, PPV and NPV of respectively 96.7%, 57.9%, 78.4%, 91.7%, highlighting CEM as a useful tool in diagnosing malignancy for its high sensitivity and in avoiding unnecessary biopsies for its high NPV [34].
In our study, of 332 ADs, 73 were malignant lesions (21.9%) and 72/73 (98.6%) of malignant ADs showed enhancement in recombined CEM images.
CEM showed a sensitivity of 98.63%, specificity of 60.62%, a PPV of 41.38% and a NPV of 99.37%, so the results from our dataset are concordant with those from previous studies, proving CEM to be a cost-effective alternative modality for the characterizations of ADs to reduce unnecessary interventional procedures. Our study suggests that a negative CEM for DBT-detected ADs may potentially allow a patient to be safely sent to clinical and radiological follow-up rather than image-guided biopsies or diagnostic surgical excision, thereby reducing the patients’ discomfort and anxiety, the cost and the possible complications related to the procedure for the healthcare system [35].
In our sample, the risk of FP results was significantly higher in case of NME vs. mass-like enhancement (p: 0.028782), but this result is not surprising since previous literature has highlighted how NME have a lower PPV for malignancy vs. mass both in CEM and MRI [36, 37].
As stated from the previous studies, our low but concordant PPV for malignancy of AD could be due to the lower PPV for DBT-only AD vs. FFDM-detected AD [38].
Also, the presence of moderate or marked BPE could have had a negative influence in the detection of breast cancer in our dataset, because of the masking effect of overlapping normal parenchymal structures, similarly to MRI [39].
Our study has several limitations, first its retrospective nature, secondly the small size of the sample with heterogeneous population including also high- and intermediate-risk women. Another important limitation is the exclusion of patients from screening programs, not performed at our institution since we are a clinical tertiary center, with potential selection bias and the fact that all ADs evaluated in our study were detected by DBT, so this could have positively biased the performance of CEM in this setting. This preliminary study aimed at validating CEM in the assessment of low-risk ADs, so we did not change the follow-up protocol, but the possibility of de-escalation is thought-provoking and will be explored in further studies given the high NPV of CEM in our results.
Lastly, CEMs were reported in consensus by three radiologists, so we did not evaluate the inter-reader agreement. However, it is worth noting that the results of our study are promising and in line with previous literature and that could benefit from a more comprehensive evaluation in a larger prospective study conducted in the future.
In conclusion, an AD without suspicious enhancement on CEM, especially in case of no or minimal BPE, has a low chance to be a breast cancer, so it can be safely sent to follow-up. Our preliminary study findings are encouraging towards the utility of CEM as a valuable complementary tool for the evaluation of AD lesions classified as BI-RADS 3, since it could contribute to the reduction of unnecessary biopsies without compromising the effectiveness of cancer detection.