Prognostic and predictive value of androgen receptor expression in postmenopausal women with estrogen receptor-positive breast cancer: results from the Breast International Group Trial 1-98.

BACKGROUND: The androgen receptor (AR) is an emerging prognostic marker and therapeutic target in breast cancer. AR is expressed in 60-80% of breast cancers, with higher prevalence among estrogen receptor-positive (ER+) tumors. Androgen treatment inhibits ER signaling in ER+/AR+ breast cancer cell lines, and AR expression is associated with improved survival for this subtype in epidemiologic studies. However, whether AR expression modifies the efficacy of selective ER modulators or aromatase inhibitors for ER+ cancers remains unclear. METHODS: We evaluated the prognostic and predictive value of AR expression among 3021 postmenopausal ER+ breast cancer patients in the Breast International Group (BIG) trial 1-98. The BIG 1-98 study was a four-armed, double-blind, phase III randomized clinical trial that compared 5 years of tamoxifen or letrozole monotherapy, or sequences of 2 years and 3 years treatment with one drug and then the other. AR expression was measured by immunohistochemistry and the percentage of AR-positive nuclei was quantified. The association between AR expression and prognosis was evaluated using Cox proportional hazards models. Continuous AR-by-treatment interactions were assessed using Subpopulation Treatment Effect Pattern Plots (STEPP). RESULTS: Eighty-two percent of patients had AR+ (≥ 1%) tumors. Patients with AR+ cancers were more likely to have smaller, lower-grade tumors, with higher expression of ER and PR. AR expression was not associated with breast cancer-free interval (BCFI) (415 events) over a median 8.0 years of follow-up (p = 0.12, log-rank test). In multivariable-adjusted models, AR expression was not associated with BCFI (HR = 1.07, 95% CI 0.83-1.36, p = 0.60). The letrozole versus tamoxifen monotherapy treatment effect did not significantly differ for AR+ tumors (HR = 0.63, 95% CI 0.44-0.75, p = 0.003) and AR- tumors (HR = 0.39, 95% CI 0.21-0.72, p = 0.002) (p-heterogeneity = 0.16). STEPP analysis also suggested no heterogeneity of the treatment effect across the continuum of AR expression. CONCLUSIONS: AR expression was not associated with prognosis, nor was there heterogeneity of the letrozole versus tamoxifen treatment effect by AR expression. These findings suggest that AR expression may not be an informative biomarker for the selection of adjuvant endocrine therapy for postmenopausal women with ER+ breast cancers. TRIAL REGISTRATION: ClinicalTrials.gov , NCT00004205, Registered 27 January 2003-Retrospectively registered, https://clinicaltrials.gov/ct2/show/study/NCT00004205 .

Reliability of a computational platform as a surrogate for manually interpreted immunohistochemical markers in breast tumor tissue microarrays.

BACKGROUND: Pathologist and computational assessments have been used to evaluate immunohistochemistry (IHC) in epidemiologic studies. We compared Definiens Tissue Studio® to pathologist scores for 17 markers measured in breast tumor tissue microarrays (TMAs) [AR, CD20, CD4, CD8, CD163, EPRS, ER, FASN, H3K27, IGF1R, IR, Ki67, phospho-mTOR, PR, PTEN, RXR, and VDR]. METHODS: 5 914 Nurses' Health Study participants, diagnosed 1976-2006 (NHS) and 1989-2006 (NHS-II), were included. IHC was conducted by the Dana-Farber/Harvard Cancer Center Specialized Histopathology Laboratory. The percent of cells staining positive was assessed by breast pathologists. Definiens output was used to calculate a weighted average of percent of cells staining positive across TMA cores for each marker. Correlations between pathologist and computational scores were evaluated with Spearman correlation coefficients. Receiver-operator characteristic curves were constructed, using pathologist scores as comparison. RESULTS: Spearman correlations between pathologist and Definiens assessments ranged from weak (RXR, rho=-0.05; CD163, rho = 0.10) to strong (Ki67, rho = 0.79; pmTOR, rho = 0.77). The area under the curve was >0.70 for all markers except RXR. CONCLUSION: Our data indicate that computational assessments exhibit variable correlations with interpretations made by an expert pathologist, depending on the marker evaluated. This study provides evidence supporting the use of computational platforms for IHC evaluation in large-scale epidemiologic studies, with the caveat that pilot studies are necessary to investigate agreement with expert assessments. In sum, computational platforms may provide greater efficiency and facilitate high-throughput epidemiologic analyses.

Deep Learning Image Analysis of Benign Breast Disease to Identify Subsequent Risk of Breast Cancer.

Background: New biomarkers of risk may improve breast cancer (BC) risk prediction. We developed a computational pathology method to segment benign breast disease (BBD) whole slide images into epithelium, fibrous stroma, and fat. We applied our method to the BBD BC nested case-control study within the Nurses' Health Studies to assess whether computer-derived tissue composition or a morphometric signature was associated with subsequent risk of BC. Methods: Tissue segmentation and nuclei detection deep-learning networks were established and applied to 3795 whole slide images from 293 cases who developed BC and 1132 controls who did not. Percentages of each tissue region were calculated, and 615 morphometric features were extracted. Elastic net regression was used to create a BC morphometric signature. Associations between BC risk factors and age-adjusted tissue composition among controls were assessed using analysis of covariance. Unconditional logistic regression, adjusting for the matching factors, BBD histological subtypes, parity, menopausal status, and body mass index evaluated the relationship between tissue composition and BC risk. All statistical tests were 2-sided. Results: Among controls, direction of associations between BBD subtypes, parity, and number of births with breast composition varied by tissue region; select regions were associated with childhood body size, body mass index, age of menarche, and menopausal status (all P < .05). A higher proportion of epithelial tissue was associated with increased BC risk (odds ratio = 1.39, 95% confidence interval = 0.91 to 2.14, for highest vs lowest quartiles, P trend = .047). No morphometric signature was associated with BC. Conclusions: The amount of epithelial tissue may be incorporated into risk assessment models to improve BC risk prediction.

Deep learning assessment of breast terminal duct lobular unit involution: Towards automated prediction of breast cancer risk.

Terminal duct lobular unit (TDLU) involution is the regression of milk-producing structures in the breast. Women with less TDLU involution are more likely to develop breast cancer. A major bottleneck in studying TDLU involution in large cohort studies is the need for labor-intensive manual assessment of TDLUs. We developed a computational pathology solution to automatically capture TDLU involution measures. Whole slide images (WSIs) of benign breast biopsies were obtained from the Nurses' Health Study. A set of 92 WSIs was annotated for acini, TDLUs and adipose tissue to train deep convolutional neural network (CNN) models for detection of acini, and segmentation of TDLUs and adipose tissue. These networks were integrated into a single computational method to capture TDLU involution measures including number of TDLUs per tissue area, median TDLU span and median number of acini per TDLU. We validated our method on 40 additional WSIs by comparing with manually acquired measures. Our CNN models detected acini with an F1 score of 0.73±0.07, and segmented TDLUs and adipose tissue with Dice scores of 0.84±0.13 and 0.87±0.04, respectively. The inter-observer ICC scores for manual assessments on 40 WSIs of number of TDLUs per tissue area, median TDLU span, and median acini count per TDLU were 0.71, 0.81 and 0.73, respectively. Intra-observer reliability was evaluated on 10/40 WSIs with ICC scores of >0.8. Inter-observer ICC scores between automated results and the mean of the two observers were: 0.80 for number of TDLUs per tissue area, 0.57 for median TDLU span, and 0.80 for median acini count per TDLU. TDLU involution measures evaluated by manual and automated assessment were inversely associated with age and menopausal status. We developed a computational pathology method to measure TDLU involution. This technology eliminates the labor-intensiveness and subjectivity of manual TDLU assessment, and can be applied to future breast cancer risk studies.

Automated clear cell renal carcinoma grade classification with prognostic significance.

We developed an automated 2-tiered Fuhrman's grading system for clear cell renal cell carcinoma (ccRCC). Whole slide images (WSI) and clinical data were retrieved for 395 The Cancer Genome Atlas (TCGA) ccRCC cases. Pathologist 1 reviewed and selected regions of interests (ROIs). Nuclear segmentation was performed. Quantitative morphological, intensity, and texture features (n = 72) were extracted. Features associated with grade were identified by constructing a Lasso model using data from cases with concordant 2-tiered Fuhrman's grades between TCGA and Pathologist 1 (training set n = 235; held-out test set n = 42). Discordant cases (n = 118) were additionally reviewed by Pathologist 2. Cox proportional hazard model evaluated the prognostic efficacy of the predicted grades in an extended test set which was created by combining the test set and discordant cases (n = 160). The Lasso model consisted of 26 features and predicted grade with 84.6% sensitivity and 81.3% specificity in the test set. In the extended test set, predicted grade was significantly associated with overall survival after adjusting for age and gender (Hazard Ratio 2.05; 95% CI 1.21-3.47); manual grades were not prognostic. Future work can adapt our computational system to predict WHO/ISUP grades, and validating this system on other ccRCC cohorts.

Establishing a Cohort of Transgender Men and Gender Nonconforming Individuals to Understand the Molecular Impact of Testosterone on Breast Physiology.

Purpose: To characterize a cohort of transgender men and masculine-centered gender nonconforming individuals who underwent gender-affirming chest-contouring surgeries at our institution between 2013 and 2018. Methods: Demographics, medical history, and breast histopathological assessment for 340 patients were retrieved from medical records. Results: Most of our patients were white, non-Hispanic (75.0%), were taking testosterone (83.2%), and opted for chest-contouring surgery after 12-14 months of testosterone therapy. Ten patients were parous (2.9%). Seventy-nine (23.2%) and 27 (7.9%) patients had a family history of breast cancer or ovarian cancer, respectively. One transgender man was incidentally diagnosed with ductal carcinoma in situ at chest-contouring surgery. Conclusion: Future studies on this cohort will provide valuable insights about the impact of testosterone on breast physiology.

PAM50 Molecular Intrinsic Subtypes in the Nurses' Health Study Cohorts.

BACKGROUND: Modified median and subgroup-specific gene centering are two essential preprocessing methods to assign breast cancer molecular subtypes by PAM50. We evaluated the PAM50 subtypes derived from both methods in a subset of Nurses' Health Study (NHS) and NHSII participants; correlated tumor subtypes by PAM50 with IHC surrogates; and characterized the PAM50 subtype distribution, proliferation scores, and risk of relapse with proliferation and tumor size weighted (ROR-PT) scores in the NHS/NHSII. METHODS: PAM50 subtypes, proliferation scores, and ROR-PT scores were calculated for 882 invasive breast tumors and 695 histologically normal tumor-adjacent tissues. Cox proportional hazards models evaluated the relationship between PAM50 subtypes or ROR-PT scores/groups with recurrence-free survival (RFS) or distant RFS. RESULTS: PAM50 subtypes were highly comparable between the two methods. The agreement between tumor subtypes by PAM50 and IHC surrogates improved to fair when Luminal subtypes were grouped together. Using the modified median method, our study consisted of 46% Luminal A, 18% Luminal B, 14% HER2-enriched, 15% Basal-like, and 8% Normal-like subtypes; 53% of tumor-adjacent tissues were Normal-like. Women with the Basal-like subtype had a higher rate of relapse within 5 years. HER2-enriched subtypes had poorer outcomes prior to 1999. CONCLUSIONS: Either preprocessing method may be utilized to derive PAM50 subtypes for future studies. The majority of NHS/NHSII tumor and tumor-adjacent tissues were classified as Luminal A and Normal-like, respectively. IMPACT: Preprocessing methods are important for the accurate assignment of PAM50 subtypes. These data provide evidence that either preprocessing method can be used in epidemiologic studies.