Siddiqui, Zafar

e14626 Background: Despite advancements in neoadjuvant therapy and surgical techniques, patients with pancreatic ductal adenocarcinoma (PDAC) often experience high recurrence rates after total neoadjuvant therapy (TNT) and resection. Existing tumor response scoring systems, such as the College of American Pathologists (CAP) guidelines, fail to stratify recurrence risk effectively, emphasizing the need for novel, objective approaches. Methods: This study evaluated spatial characteristics of residual cancer and stroma in a multi-site cohort of 203 patients with resected PDAC following TNT, not having achieved a major pathologic response (CAP 2 or 3). Whole slide images (WSI) of hematoxylin and eosin-stained sections were digitized and analyzed using artificial intelligence to perform tissue segmentation. Metrics quantifying spatial morphology and configuration, inspired by principles of landscape ecology, were computed to explore spatial features of the tumor microenvironment (TME). Statistical models (recursive partitioning and regression tree [RPRT] model and Cox model with backwards eliminations) incorporating these features were developed to stratify recurrence risk. Results: Five metrics were used for risk classification model build. Three models successfully classified patients into high- and low-risk groups based on spatial metrics. Models demonstrated moderate discriminatory power (C-statistics: 0.560–0.566) and showed that increased tumor fragmentation, like stromal patch density and cancer patch shape index, were strongly associated with improved outcomes. Conclusions: Integrating spatial metrics of cancer-stroma configuration enhances risk stratification beyond conventional scoring systems. Given increased tumor fragmentation is associated with improved outcomes, these data indirectly suggest ecological mechanisms of treatment resistance. This approach offers a pathway toward personalized post-operative management strategies to improve outcomes in PDAC. Risk classification models. Model Description Event/n Median DFS (95% CI) HR(95% CI) p-value C-statistics 1(RPRT) High risk: log standard deviation of the stroma shape index > 0.651 87 / 101 8.22(5.72 – 10.1) 1.64(1.20 – 2.25) 0.002 0.562 Low risk: log standard deviation of the stroma shape index ≤ 0.651 71 / 102 11.90(9.24 – 25.1) Ref 2(RPRT) High risk: log standard deviation of the stroma shape index > 0.651 & mean cancer shape index ≤ 1.775 58 / 63 7.23(5.19 – 9.63) 1.82(1.31 – 2.52) 0.0003 0.560 Low risk: (log standard deviation of the stroma shape index > 0.651 & mean cancer shape index > 1.775) or log standard deviation of the stroma shape index ≤ 0.651 100 / 140 11.57(9.24 – 17.92) Ref 3(Backwards elimination) High risk: PI * > median 86 / 101 8.02(6.08 – 9.86) 1.73(1.26 – 2.37) 0.0007 0.566 Low risk:PI * ≤ median 72 / 102 15.72(9.97 – 25.05) Ref *PI: (0.2701 * log mean stroma area) + (0.3905 * log edge density).

Abstract Despite advances in surgery and chemotherapy, patients with resectable pancreatic ductal adenocarcinoma (PDAC) face high recurrence rates (40-60%). Total neoadjuvant therapy (TNT) seeks to enhance margin-negative resections (R0) and address micrometastatic disease, with potential survival benefits. Yet, TNT still results in high recurrence rates, with 1-, 2-, and 3-year disease-free survival (DFS) at 65%, 48%, and 32%, respectively. Residual cancer extent is a key prognostic factor in PDAC after TNT. Tumor response scores show favorable outcomes only for patients achieving near-complete or complete pathologic response. In a cohort of 511 PDAC patients treated with TNT and surgery (2015-2022), improved DFS was only found in those with a complete pathologic response (pCR, 7.8%). This highlights the limited use of current response scoring in stratifying post-TNT patient risk, underscoring the need to better understand disease persistence and relapse. Here we developed a quantitative method to assess residual cancer. Using a digital pathology H&E library from 2180 PDAC cases (2000-2022), an AI platform segmented tissue into cancer epithelium, stroma, and normal regions. For 202 TNT-treated cases without pCR, we calculated a ratio of cancer area to stromal area from representative slides. Modeled after the histologic tumor regression grade (HTRG) system, cases were classified into <5%, 5-50%, or >50% cancer-to-stroma ratios. However, this measure alone showed no difference in DFS, suggesting tumor burden cannot fully explain differences in clinical outcomes. We hypothesized that the spatial organization of the tumor microenvironment (TME) is a determinant of patient outcomes. Drawing from landscape ecology, which studies spatial properties of ecosystems, we applied similar principles to analyze tissue architecture. Using an AI-based two-class segmentation of whole-slide images, we assessed spatial features like patch area, perimeter, density, and Shannon diversity. Python’s ‘pylandstats’ package calculated 21 spatial metrics, reduced via stepwise Cox regression using the Akaike information criterion (AIC). The final model included mean Shannon diversity, mean and standard deviation of stromal patch areas, and stromal patch density. Kaplan-Meier analysis of 202 samples stratified model predictions into quartiles, correlating with high (mean DFS 10.85 months), medium (19.32 months), and low risk (25.92 months) groups. These findings aligned with model expectations, underscoring the potential of spatial TME analysis to predict outcomes. Citation Format: Andrea V. Maraone, Zafar Siddiqui, Luis H. Cisneros, Merih D. Toruner, Mark Truty, Cornelius Thiels, Robert R. McWilliams, Steven R. Alberts, Rondell P. Graham, Carlo Maley, Qian Shi, Martin E. Fernandez-Zapico, Ryan M. Carr. Tumor spatial configuration predicts pancreatic cancer recurrence risk after total neoadjuvant therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5070.

e16368 Background: Pancreatic ductal adenocarcinoma (PDAC) is predicted to be the second cause of cancer death in the US by 2040. For the minority of patients who are candidates for curative-intent surgical resection, a neoadjuvant approach with chemotherapy or chemotherapy followed by chemoradiotherapy (chemoRT) has become the standard of care. The tumor microenvironment (TME) has been recognized as a key factor in therapeutic resistance and disease progression. This study aims to explore the prognostic value of human-interpretable machine learning image features (HIFs) derived from whole-slide H&E images. Methods: Consecutive PDAC patients who underwent neoadjuvant therapy followed by surgical resection were studied in this single-institution retrospective study. Clinical variables collected include tumor site (head/body/tail), surgery extent, histology, sex, age at diagnosis, grade, treatment response (4-point scale), LVI, PNI, resection margin status, pathological TNM staging, number of lymph nodes removed/positive, neoadjuvant strategy (chemotherapy vs chemoRT), and number of neoadjuvant chemo cycles. These factors were analyzed with a multivariable Cox proportional hazards model for disease-free survival (DFS) after surgery. Additionally, 420 human-interpretable image features (HIFs) derived from a commercial machine learning software were analyzed with a univariable Cox model on a binary outcome of 1-yr DFS, applying a Bonferroni correction for multiple comparisons to identify significant predictors. Results: Increased neoadjuvant chemotherapy cycles (HR = 0.62, p = 0.001, [median: 8, IQR: 6-12]) and receipt of combination neoadjuvant therapy (HR = 0.15, p = 0.020, [29% chemo, 71% +chemoRT]) exhibited a protective effect in the multivariable DFS model. 3 HIFs quantifying immune cells and macrophages within the cancerous tissue and its stroma were associated with an improved probability of 1-year DFS. Conclusions: In addition to clinical factors, HIFs may offer valuable insights into the tumor microenvironment's influence on PDAC progression, paving the way for improved prognostic models and therapeutic strategies. Future work will integrate clinical predictors with HIFs in a combined model.

Abstract Pancreatic ductal adenocarcinoma (PDAC) is predicted to be the second cause of cancer death in the US by 2040. Five-year overall survival is only ~12% despite decades of research. The mainstay of treatment remains chemotherapy in the absence of highly recurrent, therapeutically actionable, genetic targets. For the minority of patients (~15-20%) who are candidates for curative-intent surgical resection, six months of intensive neoadjuvant chemotherapy with/without chemoradiation has become standard-of-care. However, despite evidence of radiographic, metabolic, and biochemical responses, significant viable residual disease on pathologic analysis is almost universally observed with approximately 40-60% risk of disease recurrence. Thus, there is a critical need to better understand why PDAC remains recalcitrant to current therapeutic modalities. Pathologic evaluation of PDAC reveals a complex microenvironment consisting of neoplastic cells dynamically interacting with both “normal” cellular (e.g. mesenchymal, immune) and non-cellular (e.g. collagen) stromal elements. While biologically important, experimental models and small-scale efforts to dissect the PDAC ecosystem are limited by focus on individual tumor compartments and generation of conflicting results. Consequently, there is a need for a more comprehensive understanding of the structure and dynamics of the PDAC TME through the theoretical lens of evolutionary ecology. Current therapies are not focused on the PDAC ecosystem. We hypothesize that cell-cell interactions of PDAC cells with its local niche are important for conferring treatment resistance. Thus, better understanding of said interactions could yield novel ecologically-informed therapies. Consequently, we developed a digital pathology library of approximately 1000 PDAC resection cases (whole-slide H&E images) performed at Mayo Clinic between 2013-2021. Approximately 400 cases are treatment naive while about 600 were treated with neoadjuvant therapy. While some qualitative features of pathologic analysis have prognostic value (e.g., pathologic treatment response, presence of perineural or lymphovascular invasion, extent of desmoplasia), we aim to identify prognostic spatial relationships using tools from landscape ecology. We utilized the HoVer-Net model for nuclei segmentation and classification and spatial statistics methods and landscape fragmentation analysis to understand the spatial organization within the PDAC tumor ecology. With these methodologies, we can assess the spatial context of cell interactions, habitat/niche/microenvironment formation, ecological relationships between cell populations, and the different tissue components. The different spatial statistics/metrics are used to quantify levels of immune cell infiltration, mixing, as well as the complexity and diversity of microenvironment phenotypes leading to high levels of tumor heterogeneity. Such measurements can help predict the likelihood of a PDAC tumor adaptive potential that could lead to fast evolutionary responses to therapy, progression, and poor prognosis. Citation Format: Merih D. Toruner, Luis Cisneros, Zafar Siddiqui, Chady Meroueh, Ritambhara Singh, Carlo Maley, Martin E. Fernandez-Zapico, Ryan M. Carr. Dissecting the pancreatic cancer microenvironment using landscape ecology [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Translating Cancer Evolution and Data Science: The Next Frontier; 2023 Dec 3-6; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(3 Suppl_2):Abstract nr B026.