Development and evaluation of a prediction model for underestimated invasive breast cancer in women with ductal carcinoma in situ at stereotactic large core needle biopsy

Deep-learning model to improve histological grading and predict upstaging of atypical ductal hyperplasia / ductal carcinoma in situ on breast biopsy

AIMS

Risk stratification of atypical ductal hyperplasia (ADH) and ductal carcinoma in situ (DCIS), diagnosed using breast biopsy, has great clinical significance. Clinical trials are currently exploring the possibility of active surveillance for low-risk lesions, whereas axillary lymph node staging may be considered during surgical planning for high-risk lesions. We aimed to develop a machine-learning algorithm based on whole-slide images of breast biopsy specimens and clinical information to predict the risk of upstaging to invasive breast cancer after wide excision.

METHODS AND RESULTS

Patients diagnosed with ADH/DCIS on breast biopsy were included in this study, comprising 592 (740 slides) and 141 (198 slides) patients in the development and independent testing cohorts, respectively. Histological grading of the lesions was independently evaluated by two pathologists. Clinical information, including biopsy method, lesion size, and Breast Imaging Reporting and Data System (BI-RADS) classification of ultrasound and mammograms, were collected. Deep DCIS consisted of three deep neural networks to evaluate nuclear grade, necrosis, and stromal reactivity. Deep DCIS output comprised five parameters: total patches, lesion extent, Deep Grade, Deep Necrosis, and Deep Stroma. Deep DCIS highly correlated with the pathologists' evaluations of both slide- and patient-level labels. All five parameters of Deep DCIS were significantly associated with upstaging to invasive carcinoma in subsequent wide excisional specimens. Using multivariate logistic regression, Deep DCIS predicted upstaging to invasive carcinoma with an area under the curve (AUC) of 0.81, outperforming pathologists' evaluation (AUC, 0.71 and 0.69). After including clinical and hormone receptor status information, performance further improved (AUC, 0.87). This combined model retained its predictive power in two subgroup analyses: the first subgroup included unequivocal DCIS (excluding cases of ADH and DCIS suspicious for microinvasion) (AUC, 0.83), while the second excluded cases of high-grade DCIS (AUC, 0.81). The model was validated in an independent testing cohort (AUC, 0.81).

CONCLUSION

This study demonstrated that deep-learning models can refine histological evaluation of ADH and DCIS on breast biopsies, which may help guide future treatment planning.

High rate of postoperative upstaging of ductal carcinoma in situ when prioritizing ultrasound evaluation of mammography-detected lesions: a single-center retrospective cohort study

BACKGROUND

The initial diagnosis of ductal carcinoma in situ (DCIS) can be upstaged to invasive cancer after definitive surgery. This study aimed to identify risk factors for DCIS upstaging using routine breast ultrasonography and mammography (MG) and to propose a prediction model.

METHODS

In this single-center retrospective study, patients initially diagnosed with DCIS (January 2016-December 2017) were enrolled (final sample size = 272 lesions). Diagnostic modalities included ultrasound-guided core needle biopsy (US-CNB), MG-guided vacuum-assisted breast biopsy, and wire-localized surgical biopsy. Breast ultrasonography was routinely performed for all patients. US-CNB was prioritized for lesions visible on ultrasound. Lesions initially diagnosed as DCIS on biopsy with a final diagnosis of invasive cancer at definitive surgery were defined as "upstaged."

RESULTS

The postoperative upstaging rates were 70.5%, 9.7%, and 4.8% in the US-CNB, MG-guided vacuum-assisted breast biopsy, and wire-localized surgical biopsy groups, respectively. US-CNB, ultrasonographic lesion size, and high-grade DCIS were independent predictive factors for postoperative upstaging, which were used to construct a logistic regression model. Receiver operating characteristic analysis showed good internal validation (area under the curve = 0.88).

CONCLUSIONS

Supplemental screening breast ultrasonography possibly contributes to lesion stratification. The low upstaging rate for ultrasound-invisible DCIS diagnosed by MG-guided procedures suggests that it is unnecessary to perform sentinel lymph node biopsy for lesions invisible on ultrasound. Case-by-case evaluation of DCIS detected by US-CNB can help surgeons determine if repeating biopsy with vacuum-assisted breast biopsy is necessary or if sentinel lymph node biopsy should accompany breast-preserving surgery.

TRIAL REGISTRATION

This single-center retrospective cohort study was conducted with the approval of the institutional review board of our hospital (approval number 201610005RIND). As this was a retrospective review of clinical data, it was not registered prospectively.

Prediction of Underestimation Using Contrast-Enhanced Spectral Mammography in Patients Diagnosed as Ductal Carcinoma In Situ on Preoperative Core Biopsy

BACKGROUND

To assess the performance of contrast-enhanced spectral mammography (CESM) for the prediction of DCIS underestimation in comparison with mammography, breast US, and breast MRI.

PATIENTS AND METHODS

We prospectively enrolled patients diagnosed with DCIS on preoperative core biopsy. Visibility, lesion type, and extent on each imaging modality, CESM gray values (CGV) were evaluated. Pathologic features of core biopsy and surgery were recorded. Chi-square or Fisher's exact test were used for univariate analysis. Multivariate logistic regression analysis was used to find independent predictors for DCIS underestimation and receiver operating characteristic (ROC) curve analysis was performed.

RESULTS

A total of 113 lesions in 108 patients were analyzed (50 pure DCIS; 63 underestimated DCIS). Visibility on mammography, breast US, CESM, and breast MRI were 44%, 76%, 58%, and 80% for pure DCIS, and 73%, 81%, 86%, and 92% for underestimated DCIS. Tumor extents on surgical pathology of pure and underestimated DCIS were 1.11 ± 1.35 cm and 2.61 ± 2.09 cm. On multivariate analysis, nuclear grade and suspected invasion on core biopsy, visibility on mammography, and extent on breast MRI were independent factors for the model 1, whereas nuclear grade on core biopsy, extent on CESM, and mean CGV on MLO-recombined image were independent factors for the model 2. Area under ROC curve (AUC) was 0.843 for model 1 including breast MRI, whereas AUC was 0.823 for model 2 including CESM, which didn't show a significant difference (P = .968).

CONCLUSION

For detecting underestimated DCIS, CESM was superior to mammography and breast US, and comparable to breast MRI.

Pure Ductal Carcinoma In Situ of the Breast: Analysis of 270 Consecutive Patients Treated in a 9-Year Period

Simple Summary

Ductal carcinoma in situ (DCIS) accounts for 20 to 25% of all breast cancers and its incidence of progression to invasive ductal carcinoma is at least 13 to 50%. The aim of our retrospective observational analysis is to review the issues of this histological type of cancer. We confirmed in a wide population of 270 consecutive patients who underwent surgery in a single institute that the management of DCIS can be difficult and particularly complex. There are many variables to be taken into consideration such as the choice of the diagnostic and bioptical technique. This delicate management must be carried out in specialized centres such as Breast Units involving multiple professional figures to define and guarantee the best possible treatment for each patient.

Abstract

Introduction: Ductal carcinoma in situ (DCIS) is an intraductal neoplastic proliferation of epithelial cells that are confined within the basement membrane of the breast ductal system. This retrospective observational analysis aims at reviewing the issues of this histological type of cancer. Materials and methods: Patients treated for DCIS between 1 January 2009 and 31 December 2018 were identified from a retrospective database. The patients were divided into two groups of 5 years each, the first group including patients treated from 2009 to 2013, and the second group including patients treated from 2014 to 2018. Once the database was completed, we performed a statistical analysis to see if there were significant differences among the 2 periods. Statistical analyses were performed using GraphPad Prism software for Windows, and the level of significance was set at p < 0.05. Results: 3586 female patients were treated for breast cancer over the 9-year study period (1469 patients from 2009 to 2013 and 2117 from 2014 to 2018), of which 270 (7.53%) had pure DCIS in the final pathology. The median age of diagnosis was 59-year-old (range 36–86). In the first period, 81 (5.5%) women out of 1469 had DCIS in the final pathology, in the second, 189 (8.9%) out of 2117 had DCIS in the final pathology with a statistically significant increase (p = 0.0001). From 2009 to 2013, only 38 (46.9%) were in stage 0 (correct DCIS diagnosis) while in the second period, 125 (66.1%) were included in this stage. The number of patients included in clinical stage 0 increased significantly (p = 0.004). In the first period, 48 (59.3%) specimen margins were at a greater or equal distance than 2 mm (negative margins), between 2014 and 2018; 137 (72.5%) had negative margins. Between 2014 and 2018 the number of DCIS patients with positive margins decreased significantly (p = 0.02) compared to the first period examined. The mastectomies number increased significantly (p = 0.008) between the 2 periods, while the sentinel lymph node biopsy (SLNB) numbers had no differences (p = 0.29). For both periods analysed all the 253 patients who underwent the follow up are currently living and free of disease. We have conventionally excluded the 17 patients whose data were lost. Conclusion: The choice of the newest imaging techniques and the most suitable biopsy method allows a better pre-operative diagnosis of the DCIS. Surgical treatment must be targeted to the patient and a multidisciplinary approach discussed in the Breast Unit centres.

Predictive factors of axillary positive sentinel lymph node biopsy in extended ductal carcinoma in situ treated by simple mastectomy at once

BACKGROUND

The incidence of positive sentinel lymph node biopsy (SLNB) in ductal carcinoma in situ (DCIS) ranged from 0 to 14%. The main hypothesis would be the presence of an invasive contingent on the final histology. The objective was to identify predictive factors of sentinel lymph node positivity in the management of extended ductal carcinoma in situ treated by simple mastectomy.

METHODS

This was a retrospective study carried out at the Lorraine Cancer Institute from January 2003 to December 2017. Women with DCIS on core-needle biopsy whose management consisted of simple mastectomy and SLNB procedure were included.

RESULTS

188 patients were analyzed. Preoperatively, 18 patients (9.6%) had DCIS with microinvasion, while the others had pure DCIS. Eight patients (4.2%) had positive sentinel lymph node biopsy, the majority of which were single micrometastases. Predictive factor of node invasion was microinvasion on biopsy (p<0.01). Only in cases of pure DCIS, the percentage of positive SLNB was reduced to 2.9%. Invasive carcinoma was found in the majority of patients with positive axillary SLNB procedure (75%, n=6), compared to 16.7% (n=30) without SLNB involvement (p<0.01).

CONCLUSIONS

The low rate of positive sentinel node biopsy in pure ductal carcinoma in situ suggests that in the absence of microinvasion, the sentinel procedure would seem less appropriate. New techniques for identifying sentinel lymph node biopsy could report axillary staging after definitive histologic results.

Is sentinel lymph node biopsy indicated in patients with a diagnosis of ductal carcinoma in situ? A systematic literature review and meta-analysis

BACKGROUND

Recent discussion has suggested that some cases of ductal carcinoma in situ (DCIS) with high risk of invasive disease may require sentinel lymph node biopsy (SLNB).

METHODS

Systematic literature review identified 48 studies (9,803 DCIS patients who underwent SLNB). Separate analyses for patients diagnosed preoperatively by core sampling and patients diagnosed postoperatively by specimen pathology were conducted to determine the percentage of patients with axillary nodal involvement. Patient factors were analyzed for associations with risk of nodal involvement.

RESULTS

The mean percentage of positive SLNBs was higher in the preoperative group (5.95% vs 3.02%; P = .0201). Meta-regression analysis showed a direct association with tumor size (P = .0333) and grade (P = .00839) but not median age nor tumor upstage rate.

CONCLUSIONS

The SLNB should be routinely considered in patients with large (>2 cm) high-grade DCIS after a careful multidisciplinary discussion. In the context of breast conserving surgery, the SLNB is not routinely indicated for low- and intermediate-grade DCIS, high-grade DCIS smaller than 2 cm, or pure DCIS diagnosed by definitive surgical excision.