When I began my career in academic cancer research, I thought about preclinical models mainly in terms of biological fidelity, whether they captured the molecular features of a tumour, including relevant biomarker profiles, resistance mechanisms, and the dependencies that might make a cancer responsive to treatment. Later, in pharmaceutical oncology development, I began to see another dimension of the problem. Even a biologically sophisticated model may represent a patient population whose treatment history, and therefore whose tumour biology, has changed by the time an investigational therapy reaches the clinic.
Cancer researchers have made enormous progress in improving preclinical models. Patient-derived xenografts, organoids, metastatic models, and models derived from treatment-exposed or therapy-resistant tumours can reproduce aspects of human disease missed by traditional systems. Yet every model is generated within a therapeutic context that continues to evolve throughout drug development.
Imagine that a new cancer drug enters preclinical development today for patients whose disease has progressed after treatments A and B. Researchers select models that reflect the biology of those tumours, explore relevant resistance mechanisms, identify biomarkers, and build a compelling translational package for that setting.
From there, however, the path to patients is long. A successful programme may progress from preclinical development into first-in-human studies, later phases of clinical development, regulatory review, launch, and eventual adoption into routine practice. During those years, the treatment landscape may change repeatedly. A therapy that is experimental today may become standard treatment. A new treatment modality may move into an earlier line. A biomarker may redefine which patients receive a particular therapy, while a new combination may alter treatment sequencing altogether.
The mismatch can emerge at more than one point. By the time the investigational drug enters clinical testing, enrolled patients may already have different treatment histories from those represented in the original preclinical work. By the time the drug is approved and incorporated into routine treatment, the landscape may have shifted again, and the patients who ultimately receive it may differ further from those represented in the original models.
This matters biologically, not merely strategically. Prior therapies can impose selective pressures that favour resistant subclones, alter target expression and signaling dependencies, reshape the tumour microenvironment and immune context, and generate new mechanisms of resistance. A tumour progressing after treatments A and B may therefore have meaningfully different molecular and cellular features from one that has first encountered A, C, and B, even when both carry the same diagnosis.
For this reason, I believe temporal relevance should be considered alongside biological fidelity in preclinical oncology development.
The aim is not to predict the future or rebuild a preclinical programme whenever a new clinical result appears. It is to ask prospectively whether plausible changes in the treatment landscape could alter the biology of the patients for whom the therapy is ultimately being developed.
A practical starting point would be to identify a small number of credible future treatment scenarios informed by therapies already in late-stage clinical development, then ask which biological assumptions underlying the therapeutic hypothesis are most vulnerable to those scenarios. If an emerging therapy is likely to move earlier in the treatment sequence, should models exposed to that therapy be incorporated? Could target expression change? Could the resistance mechanism on which the new drug depends become less relevant, or perhaps more relevant?
Pharmaceutical and academic research teams already look forward when considering emerging clinical data, competing therapies, and anticipated standards of care. The opportunity is to connect that forward-looking clinical thinking more systematically back to preclinical biology.
After more than 15 years working across cancer research and oncology clinical drug development, I have come to see one principle as increasingly critical. Assumptions made during preclinical research must remain relevant as therapies progress through clinical development and into a changing treatment landscape.
When we ask whether a preclinical model represents human cancer, we are addressing only one dimension of relevance. A model can faithfully reproduce the biology of the disease it captures and still fail to reflect the biology of the patients the drug will eventually encounter. The more consequential question is whether it represents the cancer biology of the patients who will actually receive the drug years later, after the treatments that will have shaped their disease by that time.
Therefore, if we want to model the patients our cancer drugs will actually treat, we need to recognise that biological relevance is not fixed at the beginning of development. It evolves with the treatment landscape, and our preclinical thinking needs to evolve with it.
About the author
Professor David Adler, MD/PhD, MBA, is an oncology drug-development and translational medicine leader with more than 15 years of industry and academic experience. He spent a decade in senior leadership within Bayer AG’s Global Oncology Clinical Development organisation and currently serves as Chief Scientific & Medical Officer of the PATHORA Institute of Pathology & Tissue Medicine. He also holds academic appointments at the Hebrew University of Jerusalem, Ben-Gurion University of the Negev and the University of Bonn.
