Bias reduction using combined stain normalization and augmentation for AI-based classification of histological images.

Artificial intelligence (AI)-assisted diagnosis is an ongoing revolution in pathology. However, a frequent drawback of AI models is their propension to make decisions based rather on bias in training dataset than on concrete biological features, thus weakening pathologists' trust in these tools. Technically, it is well known that microscopic images are altered by tissue processing and staining procedures, being one of the main sources of bias in machine learning for digital pathology. So as to deal with it, many teams have written about color normalization and augmentation methods. However, only a few of them have monitored their effects on bias reduction and model generalizability. In our study, two methods for stain augmentation (AugmentHE) and fast normalization (HEnorm) have been created and their effect on bias reduction has been monitored. Actually, they have also been compared to previously described strategies. To that end, a multicenter dataset created for breast cancer histological grading has been used. Thanks to it, classification models have been trained in a single center before assessing its performance in other centers images. This setting led to extensively monitor bias reduction while providing accurate insight of both augmentation and normalization methods. AugmentHE provided an 81% increase in color dispersion compared to geometric augmentations only. In addition, every classification model that involved AugmentHE presented a significant increase in the area under receiving operator characteristic curve (AUC) over the widely used RGB shift. More precisely, AugmentHE-based models showed at least 0.14 AUC increase over RGB shift-based models. Regarding normalization, HEnorm appeared to be up to 78x faster than conventional methods. It also provided satisfying results in terms of bias reduction. Altogether, our pipeline composed of AugmentHE and HEnorm improved AUC on biased data by up to 21.7% compared to usual augmentations. Conventional normalization methods coupled with AugmentHE yielded similar results while being much slower. In conclusion, we have validated an open-source tool that can be used in any deep learning-based digital pathology project on H&E whole slide images (WSI) that efficiently reduces stain-induced bias and later on might help increase pathologists' confidence when using AI-based products.

Trastuzumab deruxtecan in metastatic breast cancer with variable HER2 expression: the phase 2 DAISY trial.

The mechanisms of action of and resistance to trastuzumab deruxtecan (T-DXd), an anti-HER2-drug conjugate for breast cancer treatment, remain unclear. The phase 2 DAISY trial evaluated the efficacy of T-DXd in patients with HER2-overexpressing (n = 72, cohort 1), HER2-low (n = 74, cohort 2) and HER2 non-expressing (n = 40, cohort 3) metastatic breast cancer. In the full analysis set population (n = 177), the confirmed objective response rate (primary endpoint) was 70.6% (95% confidence interval (CI) 58.3-81) in cohort 1, 37.5% (95% CI 26.4-49.7) in cohort 2 and 29.7% (95% CI 15.9-47) in cohort 3. The primary endpoint was met in cohorts 1 and 2. Secondary endpoints included safety. No new safety signals were observed. During treatment, HER2-expressing tumors (n = 4) presented strong T-DXd staining. Conversely, HER2 immunohistochemistry 0 samples (n = 3) presented no or very few T-DXd staining (Pearson correlation coefficient r = 0.75, P = 0.053). Among patients with HER2 immunohistochemistry 0 metastatic breast cancer, 5 of 14 (35.7%, 95% CI 12.8-64.9) with ERBB2 expression below the median presented a confirmed objective response as compared to 3 of 10 (30%, 95% CI 6.7-65.2) with ERBB2 expression above the median. Although HER2 expression is a determinant of T-DXd efficacy, our study suggests that additional mechanisms may also be involved. (ClinicalTrials.gov identifier NCT04132960 .).

Pattern and risk factors of isolated local relapse among women with hormone receptor-positive and HER2-negative breast cancer and lymph node involvement: 10-year follow-up analysis of the PACS 01 and PACS 04 trials.

PURPOSE: We aimed to determine the pattern of isolated local recurrences (ILR) in women with stage II-III hormone receptor-positive and human epidermal growth factor receptor 2 breast cancer (HR + /HER2-BC) after 10-year follow-up. METHODS: UNICANCER-PACS 01 and PACS 04 trials included 5,008 women with T1-T3 and N1-N3 to evaluate the efficacy of different anthracycline ± taxanes-containing regimens after modified mastectomy or lumpectomy plus axillary lymph node dissection. We analyzed the data from 2,932 women with HR + /HER2- BC to evaluate the cumulative incidence of ILR and describe the factors associated with ILR. RESULTS: After a median follow-up of 9.1 years (95% CI 9.0-9.2 years), the cumulative incidence of ILR increased steadily between 1 and 10 years from 0.2% to 2.5%. The multivariable analysis showed that older age (subhazard ratios [sHR] = 0.95, 95% CI 0.92-0.99) and mastectomy (sHR = 0.39, 95% CI 0.17-0.86) were associated with lower risk of ILR, and no adjuvant endocrine therapy (sHR = 2.73, 95% CI 1.32 7-5.67) with increased risk of ILR. CONCLUSION: In this population of high-risk patients with localized HR + /HER2- BC, the risk of ILR was low but remained constant over 10 years. Younger age at diagnosis, breast-conserving surgery, and adjuvant endocrine therapy were independent risk factors of ILR.

Uncommon invasive lobular carcinoma with papillary architecture-clinicopathologic and molecular characterization with review of the literature.

Breast cancer is the most common malignancy in woman, and its associated mortality is still rising worldwide. Among all the different subtypes of breast cancer, invasive lobular carcinoma (ILC) is the second most frequent. Several histological variants of ILC currently exist such as solid, alveolar, pleomorphic, tubulo-lobular, and mixed types. Recently, a new variant of ILC with a papillary growth pattern has been described. Here, we make a review of the literature and report the sixth case of a woman suffering from this very uncommon variant. Of note, she had a concomitant axillary lymph node metastasis, a manifestation not yet described so far. Molecular analysis showed CDH1 and PIK3CA mutations, along with similar quantitative chromosomal alterations in both primary and metastasis. Because ILC and papillary carcinoma are managed differently, our aim here is to raise awareness among the pathologists to avoid misdiagnosis of this unusual variant and subsequent inappropriate treatment.

Comparative Genomic Profiling of Second Breast Cancers following First Ipsilateral Hormone Receptor-Positive Breast Cancers.

PURPOSE: We compared the mutational profile of second breast cancers (SBC) following first ipislateral hormone receptor-positive breast cancers of patient-matched tumors to distinguish new primaries from true recurrences. EXPERIMENTAL DESIGN: Targeted next-generation sequencing using the Oncomine Tumor Mutation Load Assay. Variants were filtered according to their allele frequency ≥ 5%, read count ≥ 5X, and genomic effect and annotation. Whole genome comparative genomic hybridization array (CGH) was also performed to evaluate clonality. RESULTS: Among the 131 eligible patients, 96 paired first breast cancer (FBC) and SBC were successfully sequenced and analyzed. Unshared variants specific to the FBC and SBC were identified in 71.9% and 61.5%, respectively. Paired samples exhibited similar frequency of gene variants, median number of variants per sample, and variant allele frequency of the reported variants except for GATA3. Among the 30 most frequent gene alterations, ARIDIA, NSD2, and SETD2 had statistically significant discordance rates in paired samples. Seventeen paired samples (17.7%) exhibited common variants and were considered true recurrences; these patients had a trend for less favorable survival outcomes. Among the 8 patients with available tissue for CGH analysis and considered new primaries by comparison of the mutation profiles, 4 patients had clonally related tumors. CONCLUSIONS: Patient-matched FBC and SBC analysis revealed that only a minority of patients exhibited common gene variants between the first and second tumor. Further analysis using larger cohorts, preferably using single-cell analyses to account for clonality, might better select patients with true recurrences and thereby better inform the decision-making process.

Federated learning for predicting histological response to neoadjuvant chemotherapy in triple-negative breast cancer.

Triple-negative breast cancer (TNBC) is a rare cancer, characterized by high metastatic potential and poor prognosis, and has limited treatment options. The current standard of care in nonmetastatic settings is neoadjuvant chemotherapy (NACT), but treatment efficacy varies substantially across patients. This heterogeneity is still poorly understood, partly due to the paucity of curated TNBC data. Here we investigate the use of machine learning (ML) leveraging whole-slide images and clinical information to predict, at diagnosis, the histological response to NACT for early TNBC women patients. To overcome the biases of small-scale studies while respecting data privacy, we conducted a multicentric TNBC study using federated learning, in which patient data remain secured behind hospitals' firewalls. We show that local ML models relying on whole-slide images can predict response to NACT but that collaborative training of ML models further improves performance, on par with the best current approaches in which ML models are trained using time-consuming expert annotations. Our ML model is interpretable and is sensitive to specific histological patterns. This proof of concept study, in which federated learning is applied to real-world datasets, paves the way for future biomarker discovery using unprecedentedly large datasets.

Tissue Selection for PD-L1 Testing in Triple Negative Breast Cancer (TNBC).

Atezolizumab in combination with nab-paclitaxel has been introduced for the treatment of locally advanced or recurrent triple negative breast cancer (TNBC). Patient selection relies on the use of immunohistochemistry using a specific monoclonal PD-L1 antibody (clone SP142) in a tightly controlled companion diagnostic test (CDx) with a defined interpretative algorithm. Currently there are no standardized recommendations for selecting the optimal tissue to be tested and there is limited data to support decision making, raising the possibility that tissue selection may bias test results. We compared PD-L1 SP142 assessment in a collection of 73 TNBC cases with matched core biopsies and excision samples. There was good correlation between PD-L1-positive core biopsy and subsequent excision, but we found considerable discrepancy between PD-L1 negative core biopsy and matched excision, with a third of cases found negative on core biopsies converting to positive upon examination of the excision tissue. In view of these findings, we developed a workflow for the clinical testing of TNBC for PD-L1 and implemented it in a central referral laboratory. We present audit data from the clinical PD-L1 testing relating to 2 years of activities, indicating that implementation of this workflow results in positivity rates in our population of TNBC similar to those of IMpassion130 clinical trial. We also developed an online atlas with a precise numerical annotation to aid pathologists in the interpretation of PD-L1 scoring in TNBC.

Tumor infiltrating lymphocyte stratification of prognostic staging of early-stage triple negative breast cancer.

The importance of integrating biomarkers into the TNM staging has been emphasized in the 8th Edition of the American Joint Committee on Cancer (AJCC) Staging system. In a pooled analysis of 2148 TNBC-patients in the adjuvant setting, TILs are found to strongly up and downstage traditional pathological-staging in the Pathological and Clinical Prognostic Stage Groups from the AJJC 8th edition Cancer Staging System. This suggest that clinical and research studies on TNBC should take TILs into account in addition to stage, as for example patients with stage II TNBC and high TILs have a better outcome than patients with stage I and low TILs.

L'intelligence artificielle pourrait-elle intervenir dans l'aide au diagnostic des cancers du sein ? ­ L'exemple de HER2: Could artificial intelligence play a role in breast cancer diagnosis? ­ The example of HER2.

HER2 is an important prognostic and predictive biomarker in breast cancer. Its detection makes it possible to define which patients will benefit from a targeted treatment. While assessment of HER2 status by immunohistochemistry in positive vs negative categories is well implemented and reproducible, the introduction of a new "HER2-low" category could raise some concerns about its scoring and reproducibility. We herein described the current HER2 testing methods and the application of innovative machine learning techniques to improve these determinations, as well as the main challenges and opportunities related to the implementation of digital pathology in the up-and-coming AI era.

[2021 update of the GEFPICS' recommendations for HER2 status assessment in invasive breast cancer in France]. / Mise à jour 2021 des recommandations du GEFPICS pour l'évaluation du statut HER2 dans les cancers infiltrants du sein en France.

The last international guidelines on HER2 determination in breast cancer have been updated in 2018 by the American Society of Clinical Oncology and College of American Pathologists, on the basis of a twenty-year practice and results of numerous clinical trials. Moreover, the emerging HER2-low concept for 1+ and 2+ non amplified breast cancers lead to refine French practices for HER2 status assessment. The GEFPICS group, composed of expert pathologists, herein presents the latest French recommendations for HER2 status evaluation in breast cancer, taking into account the ASCO/CAP guidelines and introducing the HER2-low concept. In the era of personalized medicine, HER2 status assessment remains one of the most important biomarkers in breast cancer and its quality guaranties the optimal patients' care. French pathologists' commitment in theranostic biomarker quality is more than ever required to provide the most efficient cares in oncology.

Interobserver variability in the assessment of stromal tumor-infiltrating lymphocytes (sTILs) in triple-negative invasive breast carcinoma influences the association with pathological complete response: the IVITA study.

High stromal tumor-infiltrating lymphocytes (sTILs) in triple-negative breast cancer (TNBC) are associated with pathological complete response (pCR) after neoadjuvant chemotherapy (NAC). Histopathological assessment of sTILs in TNBC biopsies is characterized by substantial interobserver variability, but it is unknown whether this affects its association with pCR. Here, we aimed to investigate the degree of interobserver variability in an international study, and its impact on the relationship between sTILs and pCR. Forty pathologists assessed sTILs as a percentage in digitalized biopsy slides, originating from 41 TNBC patients who were treated with NAC followed by surgery. Pathological response was quantified by the MD Anderson Residual Cancer Burden (RCB) score. Intraclass correlation coefficients (ICCs) were calculated per pathologist duo and Bland-Altman plots were constructed. The relation between sTILs and pCR or RCB class was investigated. The ICCs ranged from -0.376 to 0.947 (mean: 0.659), indicating substantial interobserver variability. Nevertheless, high sTILs scores were significantly associated with pCR for 36 participants (90%), and with RCB class for eight participants (20%). Post hoc sTILs cutoffs at 20% and 40% resulted in variable associations with pCR. The sTILs in TNBC with RCB-II and RCB-III were intermediate to those of RCB-0 and RCB-I, with lowest sTILs observed in RCB-I. However, the limited number of RCB-I cases precludes any definite conclusions due to lack of power, and this observation therefore requires further investigation. In conclusion, sTILs are a robust marker for pCR at the group level. However, if sTILs are to be used to guide the NAC scheme for individual patients, the observed interobserver variability might substantially affect the chance of obtaining a pCR. Future studies should determine the 'ideal' sTILs threshold, and attempt to fine-tune the patient selection for sTILs-based de-escalation of NAC regimens. At present, there is insufficient evidence for robust and reproducible sTILs-guided therapeutic decisions.

Phenotypic discordance between primary and metastatic breast cancer in the large-scale real-life multicenter French ESME cohort.

Expression of hormone receptor (HR) for estrogens (ER) and progesterone (PR) and HER2 remains the cornerstone to define the therapeutic strategy for breast cancer patients. We aimed to compare phenotypic profiles between matched primary and metastatic breast cancer (MBC) in the ESME database, a National real-life multicenter cohort of MBC patients. Patients with results available on both primary tumour and metastatic disease within 6 months of MBC diagnosis and before any tumour progression were eligible for the main analysis. Among the 16,703 patients included in the database, 1677 (10.0%) had available biopsy results at MBC diagnosis and on matched primary tumour. The change rate of either HR or HER2 was 27.0%. Global HR status changed (from positive = either ER or PR positive, to negative = both negative; and reverse) in 14.2% of the cases (expression loss in 72.5% and gain in 27.5%). HER2 status changed in 7.8% (amplification loss in 45.2%). The discordance rate appeared similar across different biopsy sites. Metastasis to bone, HER2+ and RH+/HER2- subtypes and previous adjuvant endocrine therapy, but not relapse interval were associated with an HR discordance in multivariable analysis. Loss of HR status was significantly associated with a risk of death (HR adjusted = 1.51, p = 0.002) while gain of HR and HER2 discordance was not. In conclusion, discordance of HR and HER2 expression between primary and metastatic breast cancer cannot be neglected. In addition, HR loss is associated with worse survival. Sampling metastatic sites is essential for treatment adjustment.

Contemporary outcomes of metastatic breast cancer among 22,000 women from the multicentre ESME cohort 2008-2016.

AIM: Real-world data inform the outcome comparisons and help the development of new therapeutic strategies. To this end, we aimed to describe the full characteristics and outcomes in the Epidemiological Strategy and Medical Economics (ESME) cohort, a large national contemporary observational database of patients with metastatic breast cancer (MBC). METHODS: Women aged ≥18 years with newly diagnosed MBC and who initiated MBC treatment between January 2008 and December 2016 in one of the 18 French Comprehensive Cancer Centers (N = 22,109) were included. We assessed the full patients' characteristics, first-line treatments, overall survival (OS) and first-line progression-free survival, as well as updated prognostic factors in the whole cohort and among the 3 major subtypes: hormone receptor positive and HER2-negative (HR+/HER2-, n = 13,656), HER2-positive (HER2+, n = 4017) and triple-negative (n = 2963) tumours. RESULTS: The median OS of the whole cohort was 39.5 months (95% confidence interval [CI], 38.7-40.3). Five-year OS was 33.8%. OS differed significantly between the 3 subtypes (p < 0.0001) with a median OS of 43.3 (95% CI, 42.5-44.5) in HR+/HER2-; 50.1 (95% CI, 47.6-53.1) in HER2+; and 14.8 months (95% CI, 14.1-15.5) in triple-negative subgroups, respectively. Beyond performance status, the following variables had a constant significant negative prognostic impact on OS in the whole cohort and among subtypes: older age at diagnosis of metastases (except for the triple-negative subtype), metastasis-free interval between 6 and 24 months, presence of visceral metastases and number of metastatic sites ≥ 3. CONCLUSIONS: The ESME program represents a unique large-scale real-life cohort on MBC. This study highlights important situations of high medical need within MBC patients. DATABASE REGISTRATION: clinicaltrials.gov Identifier NCT032753.

Interobserver variability in upfront dichotomous histopathological assessment of ductal carcinoma in situ of the breast: the DCISion study.

Histopathological assessment of ductal carcinoma in situ, a nonobligate precursor of invasive breast cancer, is characterized by considerable interobserver variability. Previously, post hoc dichotomization of multicategorical variables was used to determine the "ideal" cutoffs for dichotomous assessment. The present international multicenter study evaluated interobserver variability among 39 pathologists who performed upfront dichotomous evaluation of 149 consecutive ductal carcinomas in situ. All pathologists independently assessed nuclear atypia, necrosis, solid ductal carcinoma in situ architecture, calcifications, stromal architecture, and lobular cancerization in one digital slide per lesion. Stromal inflammation was assessed semiquantitatively. Tumor-infiltrating lymphocytes were quantified as percentages and dichotomously assessed with a cutoff at 50%. Krippendorff's alpha (KA), Cohen's kappa and intraclass correlation coefficient were calculated for the appropriate variables. Lobular cancerization (KA = 0.396), nuclear atypia (KA = 0.422), and stromal architecture (KA = 0.450) showed the highest interobserver variability. Stromal inflammation (KA = 0.564), dichotomously assessed tumor-infiltrating lymphocytes (KA = 0.520), and comedonecrosis (KA = 0.539) showed slightly lower interobserver disagreement. Solid ductal carcinoma in situ architecture (KA = 0.602) and calcifications (KA = 0.676) presented with the lowest interobserver variability. Semiquantitative assessment of stromal inflammation resulted in a slightly higher interobserver concordance than upfront dichotomous tumor-infiltrating lymphocytes assessment (KA = 0.564 versus KA = 0.520). High stromal inflammation corresponded best with dichotomously assessed tumor-infiltrating lymphocytes when the cutoff was set at 10% (kappa = 0.881). Nevertheless, a post hoc tumor-infiltrating lymphocytes cutoff set at 20% resulted in the highest interobserver agreement (KA = 0.669). Despite upfront dichotomous evaluation, the interobserver variability remains considerable and is at most acceptable, although it varies among the different histopathological features. Future studies should investigate its impact on ductal carcinoma in situ prognostication. Forthcoming machine learning algorithms may be useful to tackle this substantial diagnostic challenge.

Improving breast cancer sensitivity to paclitaxel by increasing aneuploidy.

Predictive biomarkers for tumor response to neoadjuvant chemotherapy are needed in breast cancer. This study investigates the predictive value of 280 genes encoding proteins that regulate microtubule assembly and function. By analyzing 3 independent multicenter randomized cohorts of breast cancer patients, we identified 17 genes that are differentially regulated in tumors achieving pathological complete response (pCR) to neoadjuvant chemotherapy. We focused on the MTUS1 gene, whose major product, ATIP3, is a microtubule-associated protein down-regulated in aggressive breast tumors. We show here that low levels of ATIP3 are associated with an increased pCR rate, pointing to ATIP3 as a predictive biomarker of breast tumor chemosensitivity. Using preclinical models of patient-derived xenografts and 3-dimensional models of breast cancer cell lines, we show that low ATIP3 levels sensitize tumors to the effects of taxanes but not DNA-damaging agents. ATIP3 silencing improves the proapoptotic effects of paclitaxel and induces mitotic abnormalities, including centrosome amplification and multipolar spindle formation, which results in chromosome missegregation leading to aneuploidy. As shown by time-lapse video microscopy, ATIP3 depletion exacerbates cytokinesis failure and mitotic death induced by low doses of paclitaxel. Our results favor a mechanism by which the combination of ATIP3 deficiency and paclitaxel treatment induces excessive aneuploidy, which in turn results in elevated cell death. Together, these studies highlight ATIP3 as an important regulator of mitotic integrity and a useful predictive biomarker for a population of chemoresistant breast cancer patients.