Development of a deep-learning model tailored for HER2 detection in breast cancer to aid pathologists in interpreting HER2-low cases.

AIMS: Over 50% of breast cancer cases are "Human epidermal growth factor receptor 2 (HER2) low breast cancer (BC)", characterized by HER2 immunohistochemistry (IHC) scores of 1+ or 2+ alongside no amplification on fluorescence in situ hybridization (FISH) testing. The development of new anti-HER2 antibody-drug conjugates (ADCs) for treating HER2-low breast cancers illustrates the importance of accurately assessing HER2 status, particularly HER2-low breast cancer. In this study we evaluated the performance of a deep-learning (DL) model for the assessment of HER2, including an assessment of the causes of discordances of HER2-Null between a pathologist and the DL model. We specifically focussed on aligning the DL model rules with the ASCO/CAP guidelines, including stained cells' staining intensity and completeness of membrane staining. METHODS AND RESULTS: We trained a DL model on a multicentric cohort of breast cancer cases with HER2-IHC scores (n = 299). The model was validated on two independent multicentric validation cohorts (n = 369 and n = 92), with all cases reviewed by three senior breast pathologists. All cases underwent a thorough review by three senior breast pathologists, with the ground truth determined by a majority consensus on the final HER2 score among the pathologists. In total, 760 breast cancer cases were utilized throughout the training and validation phases of the study. The model's concordance with the ground truth (ICC = 0.77 [0.68-0.83]; Fisher P = 1.32e-10) is higher than the average agreement among the three senior pathologists (ICC = 0.45 [0.17-0.65]; Fisher P = 2e-3). In the two validation cohorts, the DL model identifies 95% [93% - 98%] and 97% [91% - 100%] of HER2-low and HER2-positive tumours, respectively. Discordant results were characterized by morphological features such as extended fibrosis, a high number of tumour-infiltrating lymphocytes, and necrosis, whilst some artefacts such as nonspecific background cytoplasmic stain in the cytoplasm of tumour cells also cause discrepancy. CONCLUSION: Deep learning can support pathologists' interpretation of difficult HER2-low cases. Morphological variables and some specific artefacts can cause discrepant HER2-scores between the pathologist and the DL model.

The Impact of COVID-19 on the Delivery of Systemic Anti-Cancer Treatment at Guy's Cancer Centre.

BACKGROUND: This study aimed to assess the outcome of cancer patients undergoing systemic anti-cancer treatment (SACT) at our centre to help inform future clinical decision-making around SACT during the COVID-19 pandemic. METHODS: Patients receiving at least one episode of SACT for solid tumours at Guy's Cancer Centre between 1 March and 31 May 2020 and the same period in 2019 were included in the study. Data were collected on demographics, tumour type/stage, treatment type (chemotherapy, immunotherapy, biological-targeted) and SARS-CoV2 infection. RESULTS: A total of 2120 patients received SACT in 2020, compared to 2449 in 2019 (13% decrease). From 2019 to 2020, there was an increase in stage IV disease (62% vs. 72%), decrease in chemotherapy (42% vs. 34%), increase in immunotherapy (6% vs. 10%), but similar rates of biologically targeted treatments (37% vs. 38%). There was a significant increase in 1st and 2nd line treatments in 2020 (68% vs. 81%; p < 0.0001) and reduction in 3rd and subsequent lines (26% vs. 15%; p = 0.004) compared to 2019. Of the 2020 cohort, 2% patients developed SARS-CoV2 infections. CONCLUSIONS: These real-world data from a tertiary Cancer Centre suggest that despite the challenges faced due to the COVID-19 pandemic, SACT was able to be continued without any significant effects on the mortality of solid-tumour patients. There was a low rate (2%) of SARS-CoV-2 infection which is comparable to the 1.4%-point prevalence in our total cancer population.

Survival Outcomes in Invasive Lobular Carcinoma Compared to Oestrogen Receptor-Positive Invasive Ductal Carcinoma.

Invasive lobular breast cancer (ILC) accounts for 10-15% of breast cancers and has distinct characteristics compared with the more common invasive ductal carcinoma (IDC). Studies have shown that ILC may be less sensitive to chemotherapy than IDC, with lower rates of complete pathological response after neo-adjuvant chemotherapy, but it is not clear how this affects long-term survival. Patients at Guy's and St Thomas' NHS Foundation Trust between 1975 and 2016 diagnosed with ER+ IDC or ER+ ILC were eligible for inclusion. Kaplan-Meier plots and Cox proportional-hazards regression models were used for analysis. There was no difference in overall survival comparing ER+ ILC to ER+ IDC (OR: 0.94, 95% CI: 0.83, 1.04) with a median follow-up time of 8.3 years compared to 8.4 years in IDC. However, ER+HER2- ILC had worse survival compared to ER+HER2- IDC in those that received chemotherapy (OR: 1.46, 95% CI: 1.06, 2.01). Here, median follow-up time was 7.0 years in ILC compared to 8.1 years in IDC. These results indicate worse overall survival after chemotherapy (neo-adjuvant and adjuvant) in ILC compared to ER+HER2- IDC even when correcting for tumour grade, age, size, and nodal involvement, but validation is needed in a larger study population.

Invasive breast cancer over four decades reveals persisting poor metastatic outcomes in treatment resistant subgroup - the "ATRESS" phenomenon.

BACKGROUND: Major advances in breast cancer treatment have led to a reducuction in mortality. However, there are still women who are not cured. We hypothesize there is a sub-group of women with treatment-resistant cancers causing early death. METHODS: Between 1975 and 2006, 5392 women with invasive breast cancer underwent surgery at Guy's Hospital, London. Data on patient demographics, tumour characteristics, treatment regimens, local recurrence, secondary metastasis, and death were prospectively recorded. We considered four time periods (1975-1982, 1983-1990, 1991-1998, 1999-2006). Risks and time to event analysis were performed with Cox proportional hazards model and Kaplan-Meier estimation. RESULTS: Unadjusted hazard ratios for developing metastasis and overall mortality relative to the 1975-1982 cohort decreased steadily to 0.23 and 0.63, respectively in 1999-2006. However, metastasis-free interval shortened, with the proportion of women developing metastasis ≤5 years increasing from 73.9% to 83.0%. Furthermore, median post-metastatic survival decreased from 1.49 years to 0.94 years. Applying our risk criteria identified the presence of ±200 patients in each cohort who developed metastasis early and died within a much shorter time frame. CONCLUSIONS: Advances in treatment have decreased the risk of metastasis and improved survival in women with invasive breast cancer over the last 40 years. Despite this, a subpopulation with shorter metastasis-free and post-metastatic survival who are unresponsive to available treatment remains. This may be due to the ATRESS phenomenon (adjuvant therapy-related shortening of survival) secondary to preselection inherent in adjuvant therapy, successful treatment of less malignant tumour cells and treatment-induced resistance in the remaining tumour clones.

Frequency of Pathogenic Germline Variants in CDH1, BRCA2, CHEK2, PALB2, BRCA1, and TP53 in Sporadic Lobular Breast Cancer.

BACKGROUND: Invasive lobular breast cancer (ILC) accounts for approximately 15% of invasive breast carcinomas and is commonly associated with lobular carcinoma in situ (LCIS). Both have been shown to have higher familial risks than the more common ductal cancers. However, there are little data on the prevalence of the known high and moderate penetrance breast cancer predisposition genes in ILC. The aim of this study was to assess the frequency of germline variants in CDH1, BRCA2, BRCA1, CHEK2, PALB2, and TP53 in sporadic ILC and LCIS diagnosed in women ages ≤60 years. METHODS: Access Array technology (Fluidigm) was used to amplify all exons of CDH1, BRCA2, BRCA1, TP53, CHEK2, and PALB2 using a custom-made targeted sequencing panel in 1,434 cases of ILC and 368 cases of pure LCIS together with 1,611 controls. RESULTS: Case-control analysis revealed an excess of pathogenic variants in BRCA2, CHEK2, PALB2, and CDH1 in women with ILC. CHEK2 was the only gene that showed an association with pure LCIS [OR = 9.90; 95% confidence interval (CI), 3.42-28.66, P = 1.4 × 10-5] with a larger effect size seen in LCIS compared with ILC (OR = 4.31; 95% CI, 1.61-11.58, P = 1.7 × 10-3). CONCLUSIONS: Eleven percent of patients with ILC ages ≤40 years carried germline variants in known breast cancer susceptibility genes. IMPACT: Women with ILC ages ≤40 years should be offered genetic screening using a panel of genes that includes BRCA2, CHEK2, PALB2, and CDH1.

Frequency of pathogenic germline variants in BRCA1, BRCA2, PALB2, CHEK2 and TP53 in ductal carcinoma in situ diagnosed in women under the age of 50 years.

INTRODUCTION: Ductal carcinoma in situ (DCIS) is a non-obligate precursor of invasive ductal breast cancer, and approximately 20% of screen-detected tumours are pure DCIS. Most risk factors for breast cancer have similar associations with DCIS and IDC; however, there is limited data on the prevalence of the known high and moderate penetrance breast cancer predisposition genes in DCIS and which women with DCIS should be referred for genetic screening. The aim of this study was to assess the frequency of germline variants in BRCA2, BRCA1, CHEK2, PALB2 and TP53 in DCIS in women aged less than 50 years of age. METHODS: After DNA extraction from the peripheral blood, Access Array technology (Fluidigm) was used to amplify all exons of these five known breast cancer predisposition genes using a custom made targeted sequencing panel in 655 cases of pure DCIS presenting in women under the age of 50 years together with 1611 controls. RESULTS: Case-control analysis revealed an excess of pathogenic variants in BRCA2 (OR = 27.96, 95%CI 6.56-119.26, P = 2.0 × 10-10) and CHEK2 (OR = 8.04, 95%CI 2.93-22.05, P = 9.0 × 10-6), with weaker associations with PALB2 (P = 0.003), BRCA1 (P = 0.007) and TP53 (P = 0.02). For oestrogen receptor (ER)-positive DCIS the frequency of pathogenic variants was 9% under the age of 50 (14% with a family history of breast cancer) and 29% under the age of 40 (42% with a family history of breast cancer). For ER-negative DCIS, the frequency was 9% (16% with a family history of breast cancer) and 8% (11% with a family history of breast cancer) under the ages of 50 and 40, respectively. CONCLUSIONS: This study has shown that breast tumourigenesis in women with pathogenic variants in BRCA2, CHEK2, PALB2, BRCA1 and TP53 can involve a DCIS precursor stage and that the focus of genetic testing in DCIS should be on women under the age of 40 with ER-positive DCIS.

Gene expression modules in primary breast cancers as risk factors for organotropic patterns of first metastatic spread: a case control study.

BACKGROUND: Metastases from primary breast cancers can involve single or multiple organs at metastatic disease diagnosis. Molecular risk factors for particular patterns of metastastic spread in a clinical population are limited. METHODS: A case-control design including 1357 primary breast cancers was used to study three distinct clinical patterns of metastasis, which occur within the first six months of metastatic disease: bone and visceral metasynchronous spread, bone-only, and visceral-only metastasis. Whole-genome expression profiles were obtained using whole genome (WG)-DASL assays from formalin-fixed paraffin-embedded (FFPE) samples. A systematic protocol was developed for handling FFPE samples together with stringent data quality controls to identify robust expression profiling data. A panel of published and novel gene sets were tested for association with these specific patterns of metastatic spread and odds ratios (ORs) were calculated. RESULTS: Metasynchronous metastasis to bone and viscera was found in all intrinsic breast cancer subtypes, while immunohistochemically (IHC)-defined receptor status and specific IntClust subgroups were risk factors for visceral-only or bone-only first metastases. Among gene modules, those related to proliferation increased the risk of metasynchronous metastasis (OR (95% CI) = 2.3 (1.1-4.8)) and visceral-only first metastasis (OR (95% CI) = 2.5 (1.2-5.1)) but not bone-only metastasis (OR (95% CI) = 0.97 (0.56-1.7)). A 21-gene module (BV) was identified in estrogen-receptor-positive breast cancers with metasynchronous metastasis to bone and viscera (area under the curve = 0.77), and its expression increased the risk of bone and visceral metasynchronous spread in this population. BV was further orthogonally validated with NanoString nCounter in primary breast cancers, and was reproducible in their matched lymph nodes metastases and an external cohort. CONCLUSION: This case-control study of WG-DASL global expression profiles from FFPE tumour samples, after careful quality control and RNA selection, revealed that gene modules in the primary tumour have differing risks for clinical patterns of metasynchronous first metastases. Moreover, a novel gene module was identified as a putative risk factor for metasynchronous bone and visceral first metastatic spread, with potential implications for disease monitoring and treatment planning.

Splicing imbalances in basal-like breast cancer underpin perturbation of cell surface and oncogenic pathways and are associated with patients' survival.

Despite advancements in the use of transcriptional information to understand and classify breast cancers, the contribution of splicing to the establishment and progression of these tumours has only recently starting to emerge. Our work explores this lesser known landscape, with special focus on the basal-like breast cancer subtype where limited therapeutic opportunities and no prognostic biomarkers are currently available. Using ExonArray analysis of 176 breast cancers and 9 normal breast tissues we demonstrate that splicing levels significantly contribute to the diversity of breast cancer molecular subtypes and explain much of the differences compared with normal tissues. We identified pathways specifically affected by splicing imbalances whose perturbation would be hidden from a conventional gene-centric analysis of gene expression. We found that a large fraction of them involve cell-to-cell communication, extracellular matrix and transport, as well as oncogenic and immune-related pathways transduced by plasma membrane receptors. We identified 247 genes in which splicing imbalances are associated with clinical patients' outcome, whilst no association was detectable at the gene expression level. These include the signaling gene TGFBR1, the proto-oncogene MYB as well as many immune-related genes such as CCR7 and FCRL3, reinforcing evidence for a role of immune components in influencing breast cancer patients' prognosis.

Family history of breast cancer and its association with disease severity and mortality.

A family history (FH) of breast cancer (BC) is known to increase an individual's risk of disease onset. However, its role in disease severity and mortality is less clear. We aimed to ascertain associations between FH of BC, severity and BC-specific mortality in a hospital-based cohort of 5354 women with prospective information on FH. We included women diagnosed at Guy's and St Thomas' NHS Foundation Trust between 1975 and 2012 (n = 5354). BC severity was defined and categorized as good, moderate, and poor prognosis. Data on BC-specific mortality was obtained from the National Cancer Registry and medical records. Associations between FH and disease severity or BC-specific mortality were evaluated using proportional odds models and Cox proportional hazard regression models, respectively. Available data allowed adjustment for potential confounders (e.g., treatment, socioeconomic status, and ethnicity). FH of any degree was not associated with disease severity at time of diagnosis (adjusted proportional OR: 1.00 [95% CI: 0.85 to 1.17]), which remained true also after stratification by period of diagnosis. FH of BC was not associated with BC-mortality HR: 0.99 (95% CI: 0.93 to 1.05). We did not find evidence to support an association between FH of BC and severity and BC-specific mortality. Our results indicate that clinical management should not differ between women with and without FH, when the underlying mutation is unknown.

Progression of breast cancer following locoregional ipsilateral recurrence: importance of interval time.

BACKGROUND: Studies comparing prognosis of breast cancer (BC) patients with and without locoregional recurrence (LR) present conflicting results. We aimed to improve our understanding of the impact of LR on prognosis by examining a large cohort of patients treated at Guy's and St Thomas' NHS Foundation Trust. METHODS: Risk factors associated with BC-specific death were investigated using Cox proportional hazards regression in 5199 women diagnosed between 1975 and 2007. Breast cancer-specific death following LR was assessed with Poisson regression. RESULTS: Overall, 552 women (11%) developed LR, with a median follow-up time of 4.28 years. Known factors associated with BC-specific death (tumour stage, grade, and nodal status) were of significance in our data. Women with a shorter disease-free interval had a worse prognosis. For instance, the HR for BC-specific death among women undergoing mastectomy with an LR 0.5-1 year after diagnosis of their primary tumour was 6.67 (95% CI: 3.71-11.99), when compared with women who did not experience LR. CONCLUSIONS: It often remains difficult to distinguish between a genuine LR and a new primary. The HRs for risk of BC-specific death following a second lesion suggest that they may act as a marker of systemic disease, large tumour burden, or depleted host defence. The clinically highly relevant impairment in prognosis calls for further research into the underlying mechanisms. We showed that for at least 15 years of follow-up, the prognosis in women following the occurrence of an LR may benefit from careful diagnostic and therapeutic management.

Integrated genomic analysis of triple-negative breast cancers reveals novel microRNAs associated with clinical and molecular phenotypes and sheds light on the pathways they control.

BACKGROUND: This study focuses on the analysis of miRNAs expression data in a cohort of 181 well characterised breast cancer samples composed primarily of triple-negative (ER/PR/HER2-negative) tumours with associated genome-wide DNA and mRNA data, extensive patient follow-up and pathological information. RESULTS: We identified 7 miRNAs associated with prognosis in the triple-negative tumours and an additional 7 when the analysis was extended to the set of all ER-negative cases. miRNAs linked to an unfavourable prognosis were associated with a broad spectrum of motility mechanisms involved in the invasion of stromal tissues, such as cell-adhesion, growth factor-mediated signalling pathways, interaction with the extracellular matrix and cytoskeleton remodelling. When we compared different intrinsic molecular subtypes we found 46 miRNAs that were specifically expressed in one or more intrinsic subtypes. Integrated genomic analyses indicated these miRNAs to be influenced by DNA genomic aberrations and to have an overall influence on the expression levels of their predicted targets. Among others, our analyses highlighted the role of miR-17-92 and miR-106b-25, two polycistronic miRNA clusters with known oncogenic functions. We showed that their basal-like subtype specific up-regulation is influenced by increased DNA copy number and contributes to the transcriptional phenotype as well as the activation of oncogenic pathways in basal-like tumours. CONCLUSIONS: This study analyses previously unreported miRNA, mRNA and DNA data and integrates these with pathological and clinical information, from a well-annotated cohort of breast cancers enriched for triple-negative subtypes. It provides a conceptual framework, as well as integrative methods and system-level results and contributes to elucidate the role of miRNAs as biomarkers and modulators of oncogenic processes in these types of tumours.

Comparison of basal-like triple-negative breast cancer defined by morphology, immunohistochemistry and transcriptional profiles.

Basal-like invasive breast cancer is an important clinical group because of its association with a triple-negative phenotype defined by the lack of expression of estrogen, progesterone and human epidermal growth factor receptors 2, relative lack of therapeutic options and poor prognosis. However, depending on the method used to define these lesions, morphological assessment, immunohistochemical markers or gene expression, a different set of tumors is captured. The aim of this study was to investigate the consequences of using different methodological approaches to define basal-like lesions among triple-negative breast carcinomas with regard to their clinicopathological features and patient outcome. The cohort consisted of 142 invasive breast cancers with a triple-negative receptor status. First, each was reviewed histologically and those with morphological basal-like features were characterized as 'Path-Basal'. Second, the 'Core Basal' immunohistochemical lesions, defined as cytokeratin 5/6 and/or epidermal growth factor receptor 1 positive, within the triple-negative breast cancers were identified, and third their classification based on gene expression profiling was retrieved and those in the molecular 'PAM50 basal-like' subtype recorded. A total of 116 basal-like breast cancers were identified among the 142 triple-negative breast cancers by at least one of these three classifications (80%), but only 13 samples were defined as basal-like with all three methods. None of these 13 tumors were associated with lymphovascular invasion. The 34 morphological 'Path-Basal' lesions were significantly associated with a lack of nodal metastases. Comparing the estimates of death in the three classifications, the highest risk of death was seen for the 'Core Basal' group. In this study, we highlight that the definition of basal-like breast cancer based on different methodologies varies significantly and does not identify the same lesions. This incomplete overlap of cases emphasizes the need for consistent or new approaches to improve precise identification.