Beyond the “Undruggable”: What Pancreatic Cancer Teaches Us About the Next Evolution of Clinical Pharmacology

Abstract

Recent reports of the pan-RAS inhibitor daraxonrasib producing substantial improvements in overall survival among patients with previously treated metastatic pancreatic ductal adenocarcinoma (PDAC) have generated unusual enthusiasm within the oncology community.^{1, 2} In the Phase III RASolute 302 study, median overall survival reportedly increased from approximately 6.7 months with standard chemotherapy to 13.2 months with daraxonrasib, a result many investigators have characterized as one of the most significant advances in pancreatic cancer treatment in decades. While not a cure, and while resistance mechanisms will undoubtedly emerge, the finding represents something more important than a successful drug. It represents the culmination of a decades-long scientific journey from a seemingly biological impossibility to therapeutic reality.

The timeline of KRAS-directed drug development is instructive. KRAS mutations were first identified in human cancers in the early 1980s and recognized as a central driver of pancreatic cancer. Yet for much of the past four decades, KRAS was considered “undruggable”.³ It is only recently that meaningful therapeutic interventions have started to become possible. During this interval, advances in structural biology enabled visualization of previously inaccessible binding pockets.⁴ Medicinal chemistry generated compounds capable of exploiting these vulnerabilities. Biomarker technologies improved patient selection. Computational methods expanded our understanding of signaling networks and resistance pathways. Clinical trial methodologies evolved to accommodate molecularly defined populations.

The lesson is not merely that scientific persistence eventually succeeded. Rather, it is that the conventional architecture of drug development was too slow, too fragmented, and too sequential for the biological complexity it sought to address. Importantly, no single innovation solved the problem. Instead, success emerged from the accumulation and integration of advances across multiple disciplines. The therapeutic breakthrough followed the maturation of an ecosystem. This observation has implications that extend well beyond pancreatic cancer.

Many diseases that remain without effective therapies today, including various neurodegenerative disorders, fibrotic diseases, rare genetic conditions, and numerous solid tumors, share characteristics with pancreatic cancer. Their biology is complex. Disease progression is heterogeneous. Therapeutic response evolves dynamically. Conventional development pathways frequently struggle to accommodate this complexity.

Recent discussions within the clinical pharmacology community have emphasized the transformative potential of artificial intelligence, quantitative systems pharmacology, model-informed drug development, advanced biomarkers, and real-world evidence.^{5, 6} These developments are often discussed as discrete innovations. Increasingly, however, their value may derive from their integration.

Historically, early-phase drug development has been organized around sequential learning. Discovery generates hypotheses. Preclinical studies evaluate mechanisms. Phase I studies establish safety and pharmacokinetics. Phase II studies explore efficacy. Phase III studies confirm clinical benefit.

This framework has served the field well for decades. Yet diseases such as pancreatic cancer expose the limitations of this framework. Biological systems evolve continuously, while development programs often acquire knowledge episodically. Molecular adaptation, clonal evolution, microenvironmental remodeling, and treatment resistance occur in real time. Our development infrastructure frequently does not.

If clinical pharmacology is to accelerate therapeutic discovery, it must evolve from a discipline focused primarily on evaluating interventions to one capable of supporting continuously adaptive learning systems.^{7, 8}

One potential future involves the creation of disease-specific translational learning ecosystems.⁹ In this model, early-phase studies would no longer function as isolated experiments. Instead, they would become components of continuously connected knowledge networks.¹⁰ Molecular profiling, pharmacokinetic data, pharmacodynamic biomarkers, imaging, circulating tumor DNA, patient-reported outcomes, and real-world clinical data would be integrated prospectively and analyzed iteratively.

The objective would not simply be to determine whether a drug works. It would be to understand why it works, when it works, for whom it works, and how resistance emerges.

Pancreatic cancer provides an ideal example. Increasing evidence suggests that response to KRAS-directed therapies is influenced not only by genotype but also by co-occurring genomic alterations, tumor microenvironment characteristics, adaptive signaling pathways, immune interactions, and treatment history.¹¹ Capturing these dynamics requires infrastructure capable of learning continuously rather than retrospectively.

Such an approach aligns naturally with advances in artificial intelligence and machine learning. Importantly, these technologies should not be viewed as replacements for scientific judgment. Rather, they serve as tools that can accelerate pattern recognition across increasingly complex datasets and identify signals that may otherwise remain undetected until much later stages of development.

The traditional objectives of Phase I studies, such as safety, tolerability, dose determination, and characterization of pharmacokinetics remain foundational. However, the future may require expanding these objectives. Increasingly, first-in-human studies should be designed to interrogate biological adaptation.

Longitudinal pharmacodynamic assessment, serial molecular profiling, digital biomarkers, quantitative systems pharmacology models, and adaptive dose optimization strategies should become routine components of development programs.¹² Rather than focusing solely on identifying a maximum tolerated dose, investigators may increasingly seek to define an optimal biologic dose,¹² an adaptive dosing strategy, or a biomarker-guided therapeutic window.

Similarly, adaptive platform trial methodologies¹³ offer opportunities to accelerate learning across multiple interventions simultaneously. By sharing infrastructure, control populations, and biomarker frameworks, such platforms may substantially reduce the time required to evaluate emerging therapeutic hypotheses. These approaches are particularly relevant for diseases with high unmet need, where every year of delay carries substantial human cost.

Perhaps the most important lesson from the pancreatic cancer story is that treatment alone may not be sufficient. Despite therapeutic advances, outcomes remain closely tied to stage at diagnosis. Consequently, future progress will require greater integration of prediction, prevention, early detection, and interception into therapeutic development paradigms.

Recent advances in multimodal risk prediction, artificial intelligence, liquid biopsy¹⁴ technologies, and longitudinal health data analysis suggest that earlier identification of high-risk individuals may become increasingly feasible. Clinical pharmacology should actively participate in validating and operationalizing these innovations.

The future of drug development may therefore extend beyond understanding how drugs interact with disease. It may increasingly involve understanding when intervention should occur and how therapeutic and diagnostic strategies can be developed together.

The recent story of KRAS and pancreatic cancer is ultimately not a story about a single drug. It is a story about the gradual construction of scientific capability. What was once considered impossible became achievable because advances in biology, chemistry, pharmacology, computation, and clinical investigation accumulated and eventually converged. The breakthrough occurred when the system became capable of learning at a level commensurate with the complexity of the disease.

This observation should inform how we think about the future of clinical pharmacology.

The next era of therapeutic innovation may not be defined primarily by better molecules, although better molecules will certainly emerge. It may be defined by better learning health systems.⁷ Systems capable of integrating biological, clinical, and digital information continuously. Systems capable of adapting in real time. Systems designed not simply to test hypotheses, but to generate them. The challenge before clinical pharmacology is ensuring that the interval between biological insight and therapeutic impact becomes dramatically shorter.

The future of drug development may depend not only on discovering the next breakthrough therapy but on building the infrastructure that allows breakthroughs to occur more frequently.

The author declares no conflicts of interest.

No funding was obtained for this work.