Can preclinical drug development predict adverse events (AE) in clinical trials?

In our last blog on preclinical drug evaluation (Preclinical safety pharmacology in a new era; Drug Discovery Today Volume 27, # 1 Jan. 2022) we discuss how conventional safety testing has not been able to identify safety issues prior the clinical phase and the novel technologies that are being generated to address this limitation (Chi et al. Drug Discovery Today 24(1) Jan 2022.

Life threatening (LT) events that have not been identified in preclinical testing are due to the lack of preclinical testing that represent human physiology.

The preclinical process is very strict but leads to a high level of attrition and resource use.

Lately, the number of therapeutic candidates reaching clinical phase I has improved (13.8%). However, this is an expensive process (average 2.6 billion USD) with a high failure rate at the clinical stage: 25% phase II, 14% phase III and 31% NDA (New drug application).

Based on this, there is a need to create methodologies that allow for the early identification of unsatisfactory safety profiles minimizing the risk for clinical studies subjects and redirection of sources (3R perspective)

Toxicity profiles not alerted in animal models, an example:

CD28 a T cell surface receptor is a therapeutic target to treat B cell chronic lymphocytic leukemia (CLL). Theralizumab, a monoclonal antibody (mAb) binds to CD28. When tested in rodents, no safety alert was issued. However, in phase I study, 6 exposed healthy volunteers suffered cytokine release syndrome (CRS) and multiorgan failure. The discrepancy explanation is the lack of CD4 effector memory in mice, so the CRS is not induced. Study in macaques fails to detect CRS due to their lack of T cell CD28. The confirmatory experiment: when human cells (stems cells, liver, or thymus) were engrafted into immunodeficient mice, and the CRS induction ability is reinstated.

Interaction between therapeutics and unintended targets: off target toxicity

Off target toxicity is a key cause of candidates´ attrition. To predict it, candidates are in vitro screened vs. targets panel known that if activated/inhibited, they cause AEs.

Effect of therapeutics on hERG or voltage-gated Potassium (K⁺) channel subfamily 2 (KCNH₂) must be prior to human studies. It has been proven that inhibition in hERG is related to QTc prolongation and to potentially fatal torsades des points (TdP).

Drug-conjugated latex/ferritin-glycidyl methacrylate beads technique isolates target proteins from cell lysates and purify/identify them in mass spectrometry. This instance could reveal potentially dangerous off-target interactions prior to clinical phase.

Metabolism

Conventional studies must evaluate in-development candidate absorption, distribution, metabolism, and excretion (ADME) stages. Cultured human hepatocytes studies use, liver microsomes, tissue slices in conjunction with animal models could reflect multi-organ physiology. However, some animal models’ relevant differences with human physiology must be considered. For example, human cytochrome enzymes behavior differs from dogs´ and rats’ microsomes not allowing them to reveal specific therapeutic toxicity.

The discrepancy is again addressed by transplanting human hepatocytes to mouse models and gathering data like that obtained in humans.

The organ on chip (OoC) methodology, is proposed to mimic human physiological environment. OoC can emulate organ function to confirm ADME data while use of multiOoC could get closer to gather human physiology. Examples of multi-OoC developed are liver-kidney and cardiac-liver systems. They have been able to respond to increase of albumin secretion with cytochrome system increase as a therapeutic exposure response or to measure cardiotoxicity via electrical/mechanical measurements on cardiomyocytes.

Microbiome role in drug toxification

It has been observed that microbiome modulations may lead to drug toxification. Nitrazepam is metabolized by microflora to an amino species that is teratogenic in rats. Antiviral (soruvidine) that has been simultaneously taken with 5-fluoruracil (FU) caused the death of 18 volunteers in a FIH study in Japan. After research on this evidence, it was concluded that the soruvidine was metabolized to bromovinylU (BVU) which irreversibly inhibited the enzyme that cleared the 5-FU resulting in a lethal accumulation.

Based on this data, there is a need to develop screening strategies to clarify the microbiome effect in the in-development candidates ADME. This need is being addressed by ex vivo fermentation platforms to incubate the candidates in human colon microbiota.

Conclusions

Whenever possible human studies should include healthy volunteers or patients with delimited medical history to avoid interference with identification of therapy induced toxicity. If drug is meant to be used chronically, in vitro, and in vivo testing should cover extended periods to increase probability of observing chronic toxicity signs.

To increase predictability of preclinical assessment, novel technologies are proposed (high throughput toxicity screening, OoC and humanized mouse models). Despite their success, these methodologies require validation: that negative findings in preclinical testing translate to a significant decrease in safety liability in the clinical setting.

As it occurred with hERG, it is widely accepted that double negative result of minimal hERG block in vitro and QTc prolongation in rodents reduces QTc increase chance from 20% to 3.8% and TdP from 10% to 0.1%. This prompted the integration of this assay to preclinical development pipelines.