Project Data Sphere

Clinical trial data analysis is grounded in rigorous statistical reasoning, employing structured methodologies to quantify uncertainty and evaluate treatment effects. With the advent of large language models (LLMs) in biomedical research, a critical question emerges: Can these models approximate expert-driven statistical analyses with accuracy and reliability? At Project Data Sphere, in collaboration with Jefferson Health and Sidney Kimmel Medical College, we put this to the test. Our study evaluated how LLMs process, interpret, and analyze clinical trial data—focusing on a study assessing a cancer therapy for extensive-stage small cell lung cancer. Key Findings: - LLMs can structure and synthesize clinical trial data effectively. - However, their statistical outputs were inconsistent, sometimes contradicting published results. - LLMs exhibited variability in safety data interpretation, occasionally overlooking critical toxicity signals. - Existing models today prioritize textual coherence over mathematical rigor, introducing risks of bias and inconsistency. Our findings highlight a fundamental limitation: LLMs lack inherent statistical constraints, rendering them inadequate for clinical trial data analysis without integration into a structured computational framework. As our CEO Sean Khozin explains, the future of AI in biostatistics isn’t about merely scaling up models but embedding them within validated statistical methodologies—leveraging Bayesian inference, probabilistic programming, and robust biostatistical principles. AI has the potential to optimize and democratize clinical trial data analysis, but its methodologies must be aligned with the foundational principles of statistical inference for interpretability. Read our preprint to explore our findings: https://lnkd.in/eWDnkAge

Clinical trial data analysis is grounded in rigorous statistical reasoning, employing structured methodologies to quantify uncertainty and evaluate treatment effects. With the advent of large language models (LLMs) in biomedical research, a critical question emerges: Can these models approximate expert-driven statistical analyses with accuracy and reliability? At Project Data Sphere, in collaboration with Jefferson Health and Sidney Kimmel Medical College, we put this to the test. Our study evaluated how LLMs process, interpret, and analyze clinical trial data—focusing on a study assessing a cancer therapy for extensive-stage small cell lung cancer. Key Findings: - LLMs can structure and synthesize clinical trial data effectively. - However, their statistical outputs were inconsistent, sometimes contradicting published results. - LLMs exhibited variability in safety data interpretation, occasionally overlooking critical toxicity signals. - Existing models today prioritize textual coherence over mathematical rigor, introducing risks of bias and inconsistency. Our findings highlight a fundamental limitation: LLMs lack inherent statistical constraints, rendering them inadequate for clinical trial data analysis without integration into a structured computational framework. As our CEO Sean Khozin explains, the future of AI in biostatistics isn’t about merely scaling up models but embedding them within validated statistical methodologies—leveraging Bayesian inference, probabilistic programming, and robust biostatistical principles. AI has the potential to optimize and democratize clinical trial data analysis, but its methodologies must be aligned with the foundational principles of statistical inference for interpretability. Read our preprint to explore our findings: https://lnkd.in/eWDnkAge

Science has never been a collection of fixed truths, only a series of evolving estimations--provisional models refined as new evidence emerges. Traditional tools in oncology, such as cancer staging, were never gold standards, only approximations shaped by the limits of their time. While these frameworks have provided structure for clinical decision-making, they rely on static classifications that fail to capture the probabilistic, fluid nature of disease. They function less as precise measurements and more as useful fictions, simplifications that have guided practice but have always been subject to revision. So where do we stand now, as AI permeates every corner of scientific and medical discourse? AI does not bring certainty, but it does bring better inference. Unlike static representations, AI-driven models can integrate clinical, molecular, and imaging data into continuously updated, probabilistic estimations of disease genesis, evolution, and burden. These models don’t impose rigid categories; they adapt to complexity, evolving in (near) real time as new data refine their predictions. Our PRECISE initiative at Project Data Sphere, for instance, replaces the flawed notion of fixed tumor staging and assessment with a dynamic, inference-driven framework, shifting how we assess progression, treatment response, and patient outcomes—not with absolutes, but with ever-improving approximations. The greatest fallacy in medicine has been mistaking structure for certainty. Traditional systems remain valuable, but they must now coexist with models that better reflect the evolving nature of disease. Science has never been a pursuit of fixed answers, only of progressively better questions. The future of precision medicine depends on embracing adaptation over finality, inference over rigid classification, and the understanding that no standard is permanent. I think the time to evolve is now-- not because AI is inevitable, but because science demands it. https://lnkd.in/dWnMPyvp

From the 2025 Nixon National Cancer Conference - "Medicine, Cancer and Artificial Intelligence" Sean Khozin, MD, MPH?– CEO, Roundtable on Cancer and Project Data Sphere Eric Stahlberg, PhD??– Former Director, Cancer Data Science Initiatives (CDSI) for Frederick National Laboratory for Cancer Research (FNLCR); Executive Administrative Director for the Institute for Data Science in Oncology at The University of Texas MD Anderson Cancer Center Clifford Hudis , MD, FACP, FASCO?(moderator) – Chief Executive Officer of, American Society of Clinical Oncology (ASCO) Watch the full panel here: https://lnkd.in/gThjSVnf

From the 2025 Nixon National Cancer Conference - "Medicine, Cancer and Artificial Intelligence" Sean Khozin, MD, MPH?– CEO, Roundtable on Cancer and Project Data Sphere Eric Stahlberg, PhD??– Former Director, Cancer Data Science Initiatives (CDSI) for Frederick National Laboratory for Cancer Research (FNLCR); Executive Administrative Director for the Institute for Data Science in Oncology at The University of Texas MD Anderson Cancer Center Clifford Hudis , MD, FACP, FASCO?(moderator) – Chief Executive Officer of, American Society of Clinical Oncology (ASCO) Watch the full panel here: https://lnkd.in/gThjSVnf

Since its inception, the CEO Roundtable on Cancer (CEORT) has been dedicated to transforming cancer research and care through strategic initiatives that span the entire innovation spectrum. From our Project Data Sphere platform advancing AI-powered analytics to our newly launched ?????????????????????? ?????? ????????????, ??????????????????????????, ?????? ?????????? (????????)?initiative focused on improving clinical trial access and diversity, CEORT brings together the most innovative minds in to accelerate progress against cancer. At the upcoming J.P. Morgan Healthcare Conference, as part of our shared commitment to modernizing cancer clinical trials and expanding access to breakthrough therapies, CEORT and Paradigm Health proudly present: ???????????????????????? ???????????? ???????????????? ????????????: ???? ?????????????????? ???????????????????? ?????????? An intimate fireside chat featuring: ?????????????? ??. ?????????????? (??????????????????) ?????????????????? ???????????????? ?????? ???????????? ??????, ?????????? & ????. ???????????? ??. ??????????????? ???????????????? ?????? ??????, ?????????? ?????????? ??. ??????????????? ?????????????????? ???????? ?????? ?????? ?????????????????? ?????????????????? ??????????????, ???????????????? ???????????? ?????????????? ???????????? ???????????????? ??????, ?????????????????? ?????????????? 14, 2025 | 2:00 ???? - 4:30 ???? ???? | ?????? ?????????????????? This forum is designed for those committed to shaping the future of cancer research and care. Senior executives interested in contributing to this important dialogue are invited to request an invitation by visiting: https://lnkd.in/e-M-tp4R ?????????? ?????????????? ???????? ???? ???????????????? ???? ?????????????????? ??????????????????. We look forward to seeing you there.

As we close an extraordinary year of progress in cancer research and care, we're pleased to share insights from our 2024 Annual Meeting—a convergence that marked a defining moment in our collective mission to improve the lives for patients with cancer. Standing at the threshold of a new era, we're reminded that the most profound breakthroughs often emerge when bold vision meets collaborative action. This year has shown us that we're at what our Board Chair Dr. David Reese calls a "hinge moment" - where advances in AI, data science, and biology are converging to fundamentally reshape our understanding of cancer. Our September Annual Meeting highlighted this transformation with the launch of PRECISE, our foundation model initiative that harnesses AI to transform medical imaging in oncology. By reimagining how we assess tumor dynamics and treatment response, we're moving beyond traditional constraints to unlock deeper insights into cancer biology. Yet technology alone isn't enough. Through our Partnership for Access, Choice, and Trust (PACT) initiative, we're tackling one of oncology's most persistent challenges: ensuring clinical trials reflect the diverse communities they aim to serve. This convergence of technological innovation and human-centered approaches exemplifies CEORT's distinctive role as a pre-competitive catalyst, where leading organizations transcend traditional boundaries to collectively advance cancer research and care, uniting scientific discovery with broader societal impact. Perhaps most inspiring is how these advances are being propelled by unprecedented collaboration. Through Project Data Sphere, we're seeing how shared data and collective intelligence can accelerate discovery in ways no single institution could achieve alone. As we reflect on 2024's achievements and look toward the promise of 2025, we're energized by the growing momentum behind our collaborative approach to conquering cancer. The future will be shaped by those who dare to reimagine what's possible and build bridges across disciplines to turn scientific promise into meaningful benefit for patients. We look forward to strengthening our partnerships and forging new connections in the year ahead, knowing that every collaboration brings us closer to the goal of eliminating the burden of cancer on patients, their families, and society. May this holiday season kindle new ideas and strengthen the bonds of collaboration that will shape tomorrow's breakthroughs.

At the upcoming J.P. Morgan Healthcare Conference, the CEO Roundtable on Cancer will discuss new initiatives that continue our tradition of uniting leaders to drive transformative solutions in oncology. Our representatives at the event include our CEO Sean Khozin, MD, MPH and Project Data Sphere President Jon McDunn, PhD. We welcome discussions with colleagues who share our commitment to precompetitive collaboration: from AI/ML innovators exploring new frontiers to academic and industry leaders driving novel research. To request a meeting, please visit: https://ceort.news/JPM2025

Project Data Sphere (PDS), a?an independent initiative of the CEO Roundtable on Cancer, is on the forefront of innovation in oncology and seeking a Chief Artificial Intelligence Officer (CAIO) to lead PDS's portfolio of AI-focused projects. Since 2019, PDS has pioneered the automation of RECIST tumor measurements and is now advancing Total Tumor Burden Index (TTBI)—a foundation model for autonomous tumor assessment. TTBI is designed to set a new standard in oncology, offering novel insights into tumor dynamics and characteristics. Why This Role Stands Out: - Help Shape the Future of Oncology: Lead the development of next-generation AI foundation models with real-world impact - Collaborate with Leaders: Work alongside top pharmaceutical companies and academic medical centers driving innovation in cancer care - Transform Tumor Detection and Assessment: Spearhead advanced in autonomous tumor detection, classification, and tracking - Impact Lives Globally: Help redefine precision oncology, delivering groundbreaking tools to improve outcomes for patients worldwide The Ideal Candidate: We’re looking for a visionary leader who blends deep technical expertise with a passion for advancing cancer care and research. This role requires not only strong AI and machine learning expertise but also a commitment to collaborative innovation in healthcare and biomedicine. Join Us: If you’re ready to make a transformative impact in oncology through the power of AI, we invite you to join this exciting journey with Project Data Sphere. Together, we can redefine what’s possible in the fight against cancer. www.CEORT.news/CAIO

This Thanksgiving, we want to express our heartfelt gratitude to our member companies, partners, and collaborators who make our mission possible. Your dedication drives meaningful progress in cancer research, care, and innovation, and we are truly thankful for your support. Through initiatives like PACT, which promotes diversity and inclusion in clinical trials, and our advances in AI and machine learning for tumor assessment and imaging analysis, we are reshaping the way cancer is diagnosed and treated. At Project Data Sphere, our shared data and collaborative platforms continue to fuel insights that bring hope to patients and families worldwide. As we reflect on our progress, we are reminded that these achievements come from the strength of our collective efforts and our shared vision of a future where the burden of cancer is lessened for everyone. Together, we are not just making strides—we are aiming to transform lives. From all of us at the CEO Roundtable on Cancer and Project Data Sphere, we wish you and your loved ones a joyful and peaceful Thanksgiving. And what better way to celebrate than with an AI-generated Thanksgiving song? May this be a season of hope and harmony! ??