Rogelio Benfield MD @GrbRoger

AI in clinical trials is no longer a marginal “optimizer” it’s starting to reshape how we design, run and interpret studies. A recent npj Digital Medicine article on “AI and innovation in clinical trials” shows how AI can refine eligibility, improve recruitment, enable adaptive designs and leverage real-world data and digital endpoints to make trials faster, more informative and more patient-centred. At the same time, it warns about opaque models driving eligibility rules, algorithmic bias worsening disparities and over-reliance on noisy realworld data without proper validation. The takeaway for oncology (and beyond) is clear: AI should enter our trials with transparency, prospective evaluation and equity as core requirements, not afterthoughts. Done right, it won’t replace rigorous clinical science it will help us generate better, fairer and more efficient evidence for our patients. nature.com/articles/s4174… #ArtificialIntelligence #ClinicalTrials #DigitalHealth #HealthcareAI #Oncology #PrecisionOncology #RealWorldData #HealthEquity

Pembrolizumab combined with platinum-based chemotherapy showed a 39.4% response rate and manageable toxicity as a potential first-line treatment for advanced penile squamous cell carcinoma. ja.ma/3XfSa8H

The new ESMO EBAI paper on AI-based biomarkers is a strong reminder that high AUC is not enough. AI tools in oncology must prove clinical applicability and non-discrimination: clear ground truth, fit-for-purpose validation (A/B/C1/C2), generalisability across centres, and fair performance in underrepresented groups. Otherwise, they’re not ready for real patients, no matter how “smart” the model looks. annalsofoncology.org/article/%20S09… #Oncology #ArtificialIntelligence #AIBiomarkers #ClinicalApplicability #NonDiscrimination #Fairness #PrecisionOncology #HealthcareAI #HealthEquity @myESMO

In an era of rapid technological advancement, physicians must embrace Machine Learning (ML) to revolutionize patient care. ML enables precise diagnostics through pattern recognition in vast datasets, facilitates personalized treatment plans, and enhances predictive analytics for disease prevention. By acquiring ML knowledge, doctors can ethically integrate AI tools, ensuring better outcomes and staying at the forefront of modern medicine. #HealthcareAI #MedTech

Just read a powerful piece in npj Artificial Intelligence on how multimodal AI is reshaping oncology. We’re no longer talking about “one more tool” but a structural redesign of how we understand, measure and treat cancer. Models that integrate genomics, imaging, clinical data and real-world outcomes let us move beyond seeing the patient as a list of variables and start understanding them as complex systems. The main bottleneck is no longer the algorithm, but fragmented data, lack of interoperability, slow regulation and the natural skepticism of clinicians facing “black boxes”. Without solid data governance, clinical metrics and alignment with payers, AI will stay as nice demos with little impact on real patients. From a Latin American perspective, the message is urgent: if we don’t build our own datasets, standards and networks, we’ll end up importing models trained on other realities and risk widening inequalities. The real revolution is not “using AI in oncology”, but redesigning oncology around multimodal, continuous, explainable and sustainable information flows. nature.com/articles/s4438… #Oncology #ArtificialIntelligence #MultimodalAI #PrecisionOncology #DigitalHealth #HealthcareAI #CancerCare #MedicalInnovation #HealthEquity #LatinAmerica @Nature_NPJ

@myESMO has just published the first framework for validating AI-based biomarkers in oncology (EBAI). It sets minimum requirements for AI systems used as biomarkers in cancer diagnosis, prognosis and treatment selection: clear ground truth, robust performance vs the gold standard, proven generalisability, plus checks for bias, explainability, cost and turnaround time. 🔗 esmo.org/society-update… #Oncology #AIinHealthcare #MedicalAI #Biomarkers #CancerResearch

Take-home for practice: •The original PACIFIC sequence (cCRT → consolidation durvalumab) remains the evidence-based standard. •Simply moving IO “earlier” in multimodality regimens doesn’t guarantee better outcomes. •Future trials need very precise thinking about where immunotherapy adds value in stage III NSCLC. Good drugs still need the right place in the algorithm. #MedTwitter #OncoTwitter #EvidenceBasedMedicine #PrecisionOncology

Results? Surprisingly neutral: •No significant PFS benefit (HR ≈ 0.85, p = 0.247) •No OS improvement (HR ≈ 1.03, p = 0.823) •Similar ORR (~61% vs 61%) and 24-month OS (~58% vs 60%) between arms Toxicity was substantial in both groups, as you’d expect with stage III cCRT, and concurrent durvalumab did not improve the efficacy–toxicity balance. #ClinicalTrials #CancerResearch #ThoracicOncology

PACIFIC-2 just reminded us that in unresectable stage III NSCLC, “earlier” immunotherapy is not automatically “better.” The trial tested starting durvalumab simultaneously with concurrent chemoradiotherapy (cCRT) and then continuing it as consolidation, vs placebo + cCRT followed by placebo. 🔗 ascopubs.org/doi/abs/10.120… #LungCancer #NSCLC #Immunotherapy #RadOnc #Oncology

The PUMP-2 phase II trial in unresectable liver-confined intrahepatic cholangiocarcinoma is hard to ignore. Dutch centers combined hepatic arterial infusion pump (HAIP) floxuridine + gemcitabine/cisplatin and reported: • 54% objective response • 86% disease control at 6 months • Median OS ≈ 22 months • 3-year OS 28–29% vs ~3% with gem-cis alone in historical ABC cohorts Single-arm, expert centers and historical comparison only — but for fit patients with liver-confined iCCA, HAIP + systemic therapy looks like a serious option to discuss in multidisciplinary boards rather than an experimental footnote. 🔗 ascopubs.org/doi/abs/10.120… #Cholangiocarcinoma #LiverCancer #GIOnc #Oncology #InterventionalOncology #CancerResearch

Just wrapped: ESMO AI & Digital Oncology Congress 2025 in Berlin. Strong focus on real-world data, multimodal AI, and how to bring trustworthy models to the clinic—not just cool demos. One to watch for future practice-changing insights. esmo.org/meeting-calend… #ESMO #DigitalOncology #AIinMedicine #OncoTwitter #Oncology

If you had to pick ONE essential AI skill for an oncologist in 2025, which would it be? Vote below and tell me why in the replies — I’m especially curious about differences between clinicians, data scientists and payers. #Oncology #AIinHealthcare #MedicalAI #OncoTwitter #MedTwitter

Radiotherapy for Unresectable Locally Advanced Non-Small Cell Lung Cancer: A Practical Multidisciplinary Approach to Challenging Scenarios from the IASLC Advanced Radiation Technology (ART) Subcommittee - Journal of Thoracic Oncology jto.org/article/S1556-…

Key challenges highlighted: • Fragmented data across silos and platforms • Lack of suitable validation datasets for ML/AI models • Articles “too technical” for clinical journals, but “not technical enough” for hard data science outlets Result: delayed translation of big-data insights into real cancer care. #MachineLearning #MedicalAI #TranslationalResearch #CancerResearch

We now have >400 cancer types, each with complex genomic, epigenomic, transcriptomic, proteomic and microenvironmental signatures. The editorial warns that the volume + complexity of these datasets are outpacing traditional structures for publishing, sharing and validating data-driven work. #PrecisionOncology #Bioinformatics #ComputationalOncology #CancerResearch

Big data is no longer a buzzword in oncology – it’s the new baseline. In Cancer Research, Christine A. Iacobuzio-Donahue reflects on how massive genomic, imaging and clinical datasets are reshaping cancer science, and why our current systems are struggling to keep up. 🔗 doi.org/10.1158/0008-5… #CancerResearch #Oncology #BigData #DataScience #AIinHealthcare #CancerResearch

⭐️RadOncRAG: A Novel Retrieval-Augmented Generation Framework Improves Large Language Model Benchmark Performance in Radiation Oncology | JCO Clinical Cancer Informatics @NGThaker_XRT @ASCO @JCOCCI_ASCO @DrArturoAI ascopubs.org/doi/10.1200/CC…

LLMs must augment, not replace oncologists. They require: •Human oversight at every level •Transparent validation in real populations •Clear regulatory and ethical frameworks If we ignore this, we don’t just risk bad AI — we risk bad oncology. #ESMO #PrecisionOncology #AIRegulation #FutureOfMedicine

Opportunities highlighted by ELCAP: ✅ Better patient education & navigation ✅ Streamlined workflows for clinicians ✅ Faster data processing for tumor boards & research But also critical risks: privacy breaches, biased outputs, hallucinations, and unsupervised use in high-stakes decisions. #MedicalAI #DataGovernance #PatientSafety

ESMO just published its first official guidance on Large Language Models in clinical oncology (ELCAP). It defines how LLMs should and should not be used in real-world cancer care — from patient chatbots to decision support and backend hospital systems. doi.org/10.1016/j.anno… #Oncology #AIinHealthcare #AIinOncology #LLM #MedTwitter #OncoTwitter