Anti-PD-1 antibodies

The mechanism, and why it complicates toxicology

These antibodies bind programmed cell death protein 1 on T cells and block its interaction with its ligand, releasing an inhibitory signal rather than acting on the tumour cell directly. That indirection is the defining feature for nonclinical work. The product has no measurable effect on a tumour cell in isolation; the readout requires a functioning immune system, and the animal must have one that the antibody can act on.

Approved products in this class

Nivolumab

Nivolumab was approved by the FDA in December 2014.^[S1] Its prescribing information is one of the two primary documents drawn on in this article, because the nonclinical section records what animal work supported the application.

Pembrolizumab

Pembrolizumab was approved in September 2014.^[S2] Some indications in this class are tied to a biomarker measured by a companion diagnostic, which is a development consideration rather than a nonclinical one, but it shapes what the nonclinical package has to support: if the intended population is defined by a marker, the models used should relate to that marker.

Cemiplimab

Cemiplimab was approved in 2018.^[S3] Three approved antibodies against the same target mean that a new entrant is generally developed either for a population the approved products do not serve or as part of a combination, and both of those change the nonclinical question from whether the mechanism works to what the specific construct or combination adds.

The problems specific to this class

Response is not uniform

Two distinct patterns are described in this class: no response from the start, and loss of an initial response. They are different problems and a model that reproduces one does not necessarily reproduce the other. The design consequence is that a study showing tumour growth inhibition in a responsive model answers a narrower question than it appears to, since the responsive setting is the one where the mechanism was least in doubt.

The tumour immune environment differs between patients

Ligand expression, mutational burden, microsatellite status and the extent of lymphocyte infiltration are all measured, and none of them alone determines the outcome. For nonclinical work the useful framing is that the immune composition of the model is itself a variable that has to be characterized and reported, not a fixed background. A syngeneic line that is heavily infiltrated and one that is not will give different answers to the same question.

Immune-related adverse reactions

The warnings sections of the approved labels in this class cover immune-mediated reactions across multiple organ systems, including pneumonitis, colitis, hepatitis and endocrinopathies.^[S1][S2] These arise from the intended pharmacology rather than from an unrelated interaction, which means they cannot be screened out by improving selectivity for the target.

The nonclinical work

Models with a functioning immune system

The standard xenograft in an immunodeficient mouse is unusable here, because the mechanism requires the T cells that model lacks. Syngeneic models supply a complete mouse immune system but require a surrogate antibody against the mouse target, so what is tested is a related molecule rather than the clinical candidate. Humanized models allow the candidate itself to be tested but carry variability between the human cell donors used to build them. Neither arrangement removes the trade-off; the choice is which limitation to accept and to state.

Measuring the immune response, not only tumour size

Because the mechanism operates through the immune system, tumour volume alone does not show whether the intended pharmacology occurred. Receptor occupancy establishes that the antibody engaged the target; flow cytometry and quantitative pathology establish what happened to the T cell populations and where they went; cytokine measurement establishes the systemic consequence. A tumour that shrank without any of these changing is a result that needs explaining before it is relied on.

Pharmacokinetics

For this class the pharmacodynamic anchor is receptor occupancy on circulating T cells rather than concentration at the tumour, since the target is on the immune cells rather than on the tumour. Occupancy tends to saturate at doses below those that produce maximal effect, so plasma concentration and biological effect do not track each other in a simple way, and dose selection cannot be read off the concentration curve alone.

Safety assessment

Anti-drug antibody formation is measured because it changes exposure within the repeat-dose study itself and can make the later timepoints uninterpretable. Cytokine release is assessed in vitro with human cells before first administration to humans. The honest position on immune-mediated organ toxicity is that animal studies in this class have limited ability to predict which organ will be affected in which patient, which is why the approved labels handle it through monitoring and management instructions rather than through exclusion.^[S1][S2]

What to establish with the laboratory

Four things are worth establishing before award. Which immune-competent models the laboratory holds and how the immune composition of each was characterized. Whether it can supply or make a mouse surrogate antibody if a syngeneic study is needed, and how the surrogate is qualified against the candidate. Whether it runs receptor occupancy assays and how the method was validated. And whether it has run non-human primate toxicology for products of this type, given that the approved products in this class were assessed that way.^[S1][S2][S4]