At the European Society for Medical Oncology (ESMO) Congress 2024, the Eyes to the Future Presidential Symposium showcased innovative approaches in personalized medicine, immunotherapy resistance, and artificial intelligence (AI)-driven pathology analysis. These presentations, focusing on the PIONeeR trial, the Rome trial, and the GigaPath AI model, respectively, offered glimpses into the future of cancer treatment and diagnosis, emphasizing the growing importance of -precision oncology and advanced technologies in improving patient outcomes.
PIONeer Trial: Overcoming PD-1/PD-L1 Inhibitor Resistance in NSCLC
Pascale Tomasini, MD, MSc
Pascale Tomasini, MD, MSc, Associate Professor in Thoracic Oncology at the Multidisciplinary Oncology and Therapeutic Innovations department in Marseille, France, presented results from the PIONeeR trial. This phase Ib/IIa clinical study aimed to address resistance to PD-1 and PD-L1 inhibitors in patients with advanced non–small cell lung cancer (NSCLC).1
This innovative trial employed an adaptive design and Bayesian statistical approach to evaluate new immunotherapy combinations against standard chemotherapy. The study enrolled 114 patients with advanced NSCLC who had experienced disease progression after previous treatment with PD-1 or PD-L1 inhibitors.
Patients were randomly assigned to receive the monoclonal antibody durvalumab in combination with various agents: monalizumab (NKG2A inhibitor), oleclumab (CD73 inhibitor), ceralasertib (ATR inhibitor), or savolitinib (MEK inhibitor). These experimental arms were compared with docetaxel chemotherapy. According to Dr. Tomasini, the trial’s adaptive design allowed for the closure of underperforming arms and the addition of new arms based on emerging data. Two arms were closed prematurely: oleclumab because of a lack of efficacy in a concurrent study and savolitinib following an interim analysis.
As Dr. Tomasini reported, the primary endpoint, the 12-week disease control rate, was not met by any of the experimental arms according to the predefined Bayesian model targets. The ceralasertib arm showed the highest disease control rate (50%) among experimental arms, comparable to docetaxel (54.5%). For monalizumab, the disease control rate was 24.1%, and with savolitinib, it was 13.6%.
In terms of objective response rates, ceralasertib demonstrated the best performance among experimental arms, with 17.9%, but this was still lower than the 25% with docetaxel. Of note, according to Dr. Tomasini, more than 40% of patients in the monalizumab arm achieved stable disease.
Although no experimental arm demonstrated formal superiority in progression-free or overall survival, Dr. Tomasini highlighted promising median overall survival durations with monalizumab (over 11 months) and ceralasertib (over 17 months), compared with 13.8 months with docetaxel.
Although some patients experienced long-term clinical benefits from the experimental combinations, said Dr. Tomasini, researchers could not identify common characteristics among these responders based on standard clinical and biological data, underscoring the need for comprehensive biomarker analyses. “Ongoing biomarker analyses will help in identifying predictive biomarkers to guide patient selection for these novel combination approaches in overcoming immunotherapy resistance in NSCLC,” Dr. Tomasini concluded.
Rome Trial: Genomic-Guided Therapy
Andrea Botticelli, MD, PhD
Andrea Botticelli, MD, PhD, Assistant Professor at Sapienza University of Rome and Coordinator of the Breast Unit at Umberto I University Hospital in Italy, presented the final results of the Rome trial. This was a randomized, multibasket phase II multicenter study exploring personalized treatment approaches based on comprehensive genomic profiling.2 Prof. Paolo Marchetti, Scientific Director of IDI IRCCS and President of the Foundation for Personalized Medicine, was the senior author of the study.
The Rome trial enrolled patients with metastatic solid tumors, regardless of histology, who had received at least one but no more than two previous lines of treatment. Comprehensive genomic profiling was performed using FoundationOne on both tissue and liquid biopsies. A molecular tumor board evaluated each case and decided on patient enrollment and treatment allocation.
From November 2020 to December 2023, 1,794 patients were screened, with 897 presenting relevant genetic alterations. After 127 weekly multidisciplinary tumor board meetings, 400 patients were enrolled and randomly assigned 1:1 to receive either the standard of care or targeted therapy based on their mutation profile.
Dr. Botticelli reported that the trial met its primary endpoint, with a significantly higher objective response rate in the targeted therapy arm compared with the standard of care (17% vs 9.5%, P < .05). The targeted therapy arm also demonstrated improved progression-free survival, with a median of 3.7 months vs 2.8 months in the standard-of-care arm (hazard ratio [HR] = 0.64, P < .05). Of note, the progression-free survival rate at 1 year was 22.3% with targeted therapy vs 7.7% with the standard of care.
Overall survival did not significantly differ between the treatment arms, with a median survival of 9.2 months with targeted therapy vs 7.6 months with the standard of care (HR = 0.89, P < .05). However, Dr. Botticelli noted, 52% of patients in the standard-of-care group crossed over to targeted therapy, potentially impacting this result.
Safety profiles were similar between treatment arms, with 35.5% of patients given targeted therapy experiencing severe toxicities compared with 40% of those given the standard of care.
A subgroup analysis of patients with high tumor mutational burden (≥ 10 mutations/megabase) given immunotherapy showed particularly promising results. In microsatellite instability–high tumors, median progression-free survival was not reached with targeted therapy vs 2.8 months with the standard of care (HR = 0.15, P < .05). In microsatellite-stable tumors, median progression-free survival was 3.6 months vs 2.8 months (HR = 0.65, P < .05).
“Targeted therapy driven by comprehensive genomic profiling and mutational tumor board evaluation significantly improved outcomes in pretreated patients with metastatic disease,” Dr. Botticelli concluded. “New trial designs are needed to expedite patient access to targeted therapies in clinical practice.”
Prov-GigaPath: Large Foundation Vision Transformers for Digital Pathology
Carlo B. Bifulco, MD
Carlo B. Bifulco, MD, Chief Medical Officer at Providence Genomics, and Medical Director of Translational Molecular Pathology and Molecular Genomics at the Earle A. Chiles Research Institute, presented groundbreaking research on the development and application of Prov-GigaPath, an open-weight, billion-parameter, whole-slide AI foundation model for cancer mutation prediction and tumor microenvironment analysis.3
“Prov-GigaPath is based on a novel vision transformer architecture and represents one of the first examples of a new class of pathology AI tools, marking a significant advancement for the field of computational pathology,” said Dr. Bifulco. “Unlike traditional convolutional neural networks, Prov-GigaPath uses a generative AI context-based approach similar to large language models and learns to predict image patches based on surrounding context, combining both image tile and whole-slide–level pretraining.” Furthermore, according to Dr. Bifulco, this allows the models to learn fundamental features from images without the need for explicit labeling, enhancing the robustness and generalizability of the models across different applications.
The Prov-GigaPath model was trained on an extensive data set of 170,000 pathology slides and 1.3 billion image tiles and incorporates not only image data but also genomic data sets and text from pathology and clinical reports. Dr. Bifulco highlighted Prov-GigaPath model’s ability to predict genomic alterations from pathology slides, potentially democratizing access to personalized medicine and clinical trials. For example, the model demonstrated superior performance in predicting EGFR mutations compared with convolutional neural network–based models.
Of note, said Dr. Bifulco, the Prov-GigaPath model has been released as an open-weight model and has already accrued more than 140,000 downloads worldwide, since its release in May 2024. Independent benchmarks have shown its high performance in tasks beyond those for which it was specifically trained.
“As pathology evolves to include complex, multidimensional spatial data sets that map conventional medical imaging and highly multiplexed biomarkers such as proteins, RNA expression, and methylation states, AI models such as Prov-GigaPath could represent a new tool in interpreting and deriving biological insights from these complex multimodal data,” Dr. Bifulco concluded.”
DISCLOSURE: Dr. Tomasini reported financial relationships with Roche, AstraZeneca, BMS, Lilly, Janssen, Amgen, and Takeda. Dr. Botticelli has served as an advisor or consultant for Eli Lilly, Pfizer, Novartis, Roche, BMS, AstraZeneca, MSD, Daiichi Sankyo, Gilead Sciences, and Seagen. Dr. Bifulco reported financial relationships with PrimeVax, BioAI, SeronaDx, Lunaphore, Sanofi, Roche, Agilent, Incendia, and Illumina.
REFERENCES
1. Tomasini P, Cropet C, Jeanson A, et al: Precision immuno-oncology for advanced non-small cell lung cancer patients with PD-(L)1 inhibitors resistance (PIONeeR). ESMO Congress 2024. Abstract LBA8. Presented September 16, 2024.
2. Botticelli A, Scagnoli S, Conte P, et al: The Rome trial from histology to target: The road to personalize targeted therapy and immunotherapy. ESMO Congress 2024. Abstract LBA7. Presented September 16, 2024.
3. Bifulco C, Poon H, Usuyama N, et al: Application of GigaPath: An open-weight billion-parameter AI foundation model based on a novel vision transformer architecture for cancer mutation prediction and TME analysis. ESMO Congress 2024. Abstract 1942O. Presented September 16, 2024.
