Detection of ERBB2 amplification in uterine serous carcinoma by next-generation sequencing: an approach highly concordant with standard assays.

Uterine serous carcinoma is an aggressive subtype of endometrial cancer that accounts for fewer than 10% of endometrial carcinomas but is responsible for about half of deaths. A subset of cases has HER2 overexpression secondary to ERBB2 gene amplification, and these patients may benefit from anti-HER2 therapies, such as trastuzumab. HER2 protein overexpression is currently assessed by immunohistochemistry (IHC) and ERBB2 gene amplification by fluorescence in situ hybridization (FISH). Targeted next-generation sequencing (NGS) is increasingly used to routinely identify predictive and prognostic molecular abnormalities in endometrial carcinoma. To investigate the ability of a targeted NGS panel to detect ERBB2 amplification, we identified cases of uterine serous carcinoma (n = 93) and compared HER2 expression by IHC and copy number assessed by FISH with copy number status assessed by NGS. ERBB2 copy number status using a combination of IHC and FISH was interpreted using the 2018 ASCO/CAP guidelines for breast carcinoma. ERBB2 amplification by NGS was determined by the relative number of reads mapping to ERBB2 in tumor DNA compared to control nonneoplastic DNA. Cases with copy number ≥6 were considered amplified and copy number <6 were non-amplified. By IHC, 70 specimens were classified as negative (0 or 1+), 19 were classified as equivocal (2+), and 4 were classified as positive (3+). Using combined IHC/FISH, ERBB2 amplification was observed in 8 of 93 cases (9%). NGS identified the same 8 cases with copy number ≥6; all 85 others had copy number <6. In this series, NGS had 100% concordance with combined IHC/FISH in identifying ERBB2 amplification. NGS is highly accurate in detecting ERBB2 amplification in uterine serous carcinoma and provides an alternative to measurement by IHC and FISH.

Assessment of Interobserver Agreement Among Gynecologic Pathologists Between Three-Tier Versus Binary Pattern-based Classification Systems for HPV-associated Endocervical Adenocarcinoma.

The three-tier (A vs. B vs. C) pattern-based (Silva) classification system is a strong and fairly reproducible predictor of the risk of lymph node involvement and recurrence of human papillomavirus (HPV)-associated endocervical adenocarcinoma (EA). Recently, a binary pattern-based classification system has been proposed which incorporates the Silva pattern and lymphovascular invasion (LVI) to assign tumors as "low risk" or "high risk" and this may have superior prognostic significance compared with the three-tier system as well as current International Federation of Gynecology and Obstetrics (FIGO) staging of cervix-confined disease. The interobserver reproducibility of this binary system, however, is unknown. Representative slides from 59 HPV-associated EAs (1-3 slides/case) were independently reviewed by 5 gynecologic pathologists who participated in an online training module before the study. In the first review, a pattern was assigned using the three-tier system. On the second review, a "low risk" or "high risk" designation was assigned and the presence or absence of LVI was specifically documented. Interobserver agreement was assessed using Fleiss' kappa. The binary system showed improved interobserver agreement (kappa=0.634) compared with the three-tier system (kappa=0.564), with a higher proportion of cases having agreement between at least 4/5 reviewers (86% vs. 73%). Nineteen and 8 cases showed improved and worse interobserver agreement using the binary system, respectively; the remainder showed no change. 3/5 reviewers showed no intraobserver discrepancy while the remaining 2 did in a small subset of cases (n=2 and 4, respectively). In this study, a binary pattern-based classification system showed improved interobserver agreement compared with the traditional three-tier system.

Deep learning models for histologic grading of breast cancer and association with disease prognosis.

Histologic grading of breast cancer involves review and scoring of three well-established morphologic features: mitotic count, nuclear pleomorphism, and tubule formation. Taken together, these features form the basis of the Nottingham Grading System which is used to inform breast cancer characterization and prognosis. In this study, we develop deep learning models to perform histologic scoring of all three components using digitized hematoxylin and eosin-stained slides containing invasive breast carcinoma. We first evaluate model performance using pathologist-based reference standards for each component. To complement this typical approach to evaluation, we further evaluate the deep learning models via prognostic analyses. The individual component models perform at or above published benchmarks for algorithm-based grading approaches, achieving high concordance rates with pathologist grading. Further, prognostic performance using deep learning-based grading is on par with that of pathologists performing review of matched slides. By providing scores for each component feature, the deep-learning based approach also provides the potential to identify the grading components contributing most to prognostic value. This may enable optimized prognostic models, opportunities to improve access to consistent grading, and approaches to better understand the links between histologic features and clinical outcomes in breast cancer.

Determining breast cancer biomarker status and associated morphological features using deep learning.

Background: Breast cancer management depends on biomarkers including estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 (ER/PR/HER2). Though existing scoring systems are widely used and well-validated, they can involve costly preparation and variable interpretation. Additionally, discordances between histology and expected biomarker findings can prompt repeat testing to address biological, interpretative, or technical reasons for unexpected results. Methods: We developed three independent deep learning systems (DLS) to directly predict ER/PR/HER2 status for both focal tissue regions (patches) and slides using hematoxylin-and-eosin-stained (H&E) images as input. Models were trained and evaluated using pathologist annotated slides from three data sources. Areas under the receiver operator characteristic curve (AUCs) were calculated for test sets at both a patch-level (>135 million patches, 181 slides) and slide-level (n = 3274 slides, 1249 cases, 37 sites). Interpretability analyses were performed using Testing with Concept Activation Vectors (TCAV), saliency analysis, and pathologist review of clustered patches. Results: The patch-level AUCs are 0.939 (95%CI 0.936-0.941), 0.938 (0.936-0.940), and 0.808 (0.802-0.813) for ER/PR/HER2, respectively. At the slide level, AUCs are 0.86 (95%CI 0.84-0.87), 0.75 (0.73-0.77), and 0.60 (0.56-0.64) for ER/PR/HER2, respectively. Interpretability analyses show known biomarker-histomorphology associations including associations of low-grade and lobular histology with ER/PR positivity, and increased inflammatory infiltrates with triple-negative staining. Conclusions: This study presents rapid breast cancer biomarker estimation from routine H&E slides and builds on prior advances by prioritizing interpretability of computationally learned features in the context of existing pathological knowledge.

Validation of the MedUseQ: A Self-Administered Screener for Older Adults to Assess Medication Use Problems.

BACKGROUND: Older adults taking multiple prescription and nonprescription drugs are at risk for medication use problems, yet there are few brief, self-administered screening tools designed specifically for them. OBJECTIVE: The study objective was to develop and validate a patient-centered screener for community-dwelling older adults. METHODS: In phase 1, a convenience sample of 57 stakeholders (older adults, pharmacists, nurses, and physicians) participated in concept mapping, using Concept System® Global MAXTM, to identify items for a questionnaire. In phase 2, a 40-item questionnaire was tested with a convenience sample of 377 adults and a 24-item version was tested with 306 older adults, aged 55 and older, using Rasch methodology. In phase 3, stakeholder focus groups provided feedback on the format of questionnaire materials and recommended strategies for addressing problems. RESULTS: The concept map contained 72 statements organized into 6 conceptual clusters or domains. The 24-item screener was unidimensional. Cronbach's alpha was .87, person reliability was acceptable (.74), and item reliability was high (.96). CONCLUSION: The MedUseQ is a validated, patient-centered tool targeting older adults that can be used to assess a wide range of medication use problems in clinical and community settings and to identify areas for education, intervention, or further assessment.