Margaret S Brandwein-Weber, MD

"Dr. Brandwein-Weber is known for the development and validation of a predictive model for patients with head and neck squamous cell carcinoma (SCC) referred to as the Risk Model, which has been validated in a multi-center study of nearly 300 patients with Stage I/II oral cavity SCC.1 Presently, it is the only validated model that can predict treatment failure for patients with Stage I/II oral cavity SCC. This can potentially shift treatment paradigms as low-stage/high-risk patients could be offered adjuvant radiotherapy. The model is based on assessing resection specimens for three significant histologic variables: tumor worst pattern of invasion (WPOI), perineural invasion, and local immune response; point assignment allows for tumor classification as either low-, intermediate- or high-risk. WPOI is stratified into “nonaggressive” versus “aggressive”. The former group encompasses patterns historically recognized as WPOI-1 through WPOI-3. Aggressive WPOI notably recognizes and splits out dispersed tumors (WPOI-5) from nondispersed tumors (WPOI-4). WPOI-4 is defined as having small tumor satellites convincingly separated from the main tumor mass, whereas WPOI-5 has the additional feature of dispersed satellites ≥ 1mm from the main tumor mass or other satellites. The presence of WPOI-5 automatically qualifies for classification as high-risk. With respect to predicting locoregional recurrence, WPOI-5 has a positive predictive value of 42%. The Risk Model has already impacted head and neck oncology as it is discussed in the 8th American Joint Committee on Cancer staging manual for oral cavity cancers. The rationale for paring down the assessment from the multivariable Risk Model to assessing for the presence or absence of WPOI-5 is that a simpler tool is more likely to be generally adapted for use. University of Alabama at Birmingham research (2010-2015) focused on tumor invasion, providing a novel biologic rationale for WPOI. Primary cancer cell lines were generated from surgical resections of tumors with different WPOI phenotypes (aggressive versus nonaggressive). Multiple invasion assays in different formats confirmed that tumor cell lines derived from nonaggressive WPOI tumors are less invasive than those derived from aggressive WPOI tumors, as would have been predicted from the phenotypic differences. Upregulating and silencing experiments demonstrated that this invasiveness is directly related to tumor suppressor protein p16(INK4a), impacting downstream regulation of extracellular matrix remodeling.2 The State University of New York at Buffalo, Department of Pathology and Anatomical Sciences has been at the forefront of quantitative digital pathology. A large body of evidence has demonstrated the benefit of digitization and computational analysis of glass slides. The process of pathologic interpretation of resection specimen slides includes generating a mental translation of planar cross-sectional 2D imagery into3D tissue architecture. In evaluating WPOI, the decision point is whether or not a tumor satellite is truly separated from the main tumor mass. Tangential tissue sectioning through a contiguous structure is known to mimic separation when viewed in cross section. Thus, a series of visual context-dependent assumptions are necessary to this interpretive process. The effectiveness of WPOI interpretation may be enhanced by building 3D models of tumor interface. We have developed a system to generate 3D spatial tumor models from traditional digitalized serial slide sections. A supervised classifier automatically generates a map of tumor regions on each digitalized slide, which are then co-registered using algorithms, and a smooth 3D model of the tumor is generated from the registered maps. We expect that a robust modeling and quantification approach to 3D histology will lead to developing a comprehensive model of disease derived from imaging data. Spatial models can also provide a structural framework upon which cellular functional markers can be mapped. Just as generating 3D models enhances the information abstracted from planar tissue sections, we believe that the additional mapping of functional biomarkers onto 3D interfaces (e.g. cytotoxic lymphocytes at the tumor/host interface) will further enhance understanding of disease and predictive modeling. The current standard of care for all Stage I/II patients with oral cavity SCC remains single modality surgery. Subgroup analysis of data support the idea that adjuvant radiotherapy is beneficial for patients with Stage I/II / WPOI-5 tumors. Thus, we continue to accrue the necessary number of Stage I/II / WPOI-5 patients to reach statistical significance, as an observational study. These data will provide the rationale for a prospective clinical trial comparing outcomes for Stage I/II / WPOI-5 patients treated by surgery alone, versus surgery plus adjuvant radiotherapy. Lastly, WPOI categories, like all pathologist-driven classifiers, originates from discrete, logical and intuitive histological variables. A limitation of this approach is the inability to test for continuous variables. We are using a graphic-based approach for extracting and quantifying architectural features of tumor silhouettes derived from WPOI-4 and WPOI-5 oral cavity SCC (“mugshots”) for the purposes of predicting disease progression. "