Nonclinical safety studies
Nonclinical safety studies test a candidate medicine in vitro and in animals to reduce the risk of undesired effects when people are exposed to it, and they run from early discovery through clinical development and sometimes beyond approval. The studies a programme needs depend on the kind of molecule and are set by a family of guidelines; the pivotal ones are run under good laboratory practice; species and doses have to be chosen; each finding has to be interpreted for what it means for humans; and the package is translated into the starting dose of the first trial in humans.
Preclinical and nonclinical
Preclinical development is one part of nonclinical development, the part that happens before first-in-human studies; nonclinical development continues past that point, up to the new drug application, marketing authorisation application or biologics licence application, with other types of study and with extensions of the general toxicology studies performed in the preclinical part.
The term preclinical assumes that all such studies are completed before a drug is first given to humans, whereas most of them are conducted at the same time as the clinical trials, so nonclinical is the more strictly correct term. Many people assume that preclinical studies are conducted and completed before clinical trials are initiated, which reflects a common view that preclinical development is of little importance once human data have been obtained, and the purpose usually given for it is that it mostly consists of toxicology studies run to ensure a compound's safety before testing in humans. Traditionally, preclinical evaluation has been viewed as an intermediate and isolated step between drug discovery and clinical development. Two nonclinical regulatory writers hold instead that it is a comprehensive scientific programme spanning the whole lifecycle of the compound, covering domains from the biological research underpinning its mode of action through its disposition in the body to safety assessment, and they use nonclinical to underscore this.
What the studies are for
The programme exists to reduce the risk of undesired effects in humans by testing compounds in vitro or in animals before people are exposed, and it scales with the exposure being enabled: the longer the clinical study, the more nonclinical work is needed to support it. The programme also identifies the specific parameters that should be monitored in the clinical studies, gives some information on the potential for drug–drug interactions, and provides the data that enable first-in-human and later human studies, including the safe maximum exploratory doses that can be used in a trial. Some studies are run because risk or ethics make it impossible to study a specific effect in humans, which can relate to the compound's mechanism or adverse biological characteristics.
Preclinical studies aim to provide information about the safety and efficacy of a candidate before it is tested in humans, can provide evidence of its biological effect, and usually include both in vitro and in vivo work; authorities such as the FDA require them before an investigational new drug filing. Insight into the compound's dosing and toxicity levels, which the pharmacokinetic, pharmacodynamic and toxicology studies provide, is essential to determine whether it is justified and reasonably safe to proceed to clinical studies.
When the studies run
In practice nonclinical studies start during the discovery phase and continue until, and sometimes beyond, approval. Nonclinical evaluation has no definitive starting point, unlike the clinical programme, where approval of the first clinical trial marks the beginning. The start of the pivotal good-laboratory-practice toxicity studies is a significant milestone, but it is preceded by many investigations reaching back into early discovery which the guidelines do not specifically mandate and which scientific and practical reasoning drives: no company would run an expensive GLP toxicity study without extensive screening for possible toxicities in silico, in vitro and in smaller-scale non-GLP in vivo work, and early assessment of pharmacokinetic properties informs both the discovery programme, as in compound optimisation, and the design of the safety evaluations, as in safety-relevant human metabolites. If the compound progresses, almost all of those early studies become part of the nonclinical package.
Although many nonclinical studies are completed before the first human trials, nonclinical development continues far beyond them. Some later studies support the late clinical phases, chronic repeated-dose toxicity among them, and others are required for marketing approval, such as carcinogenicity; planning them to run in parallel with clinical development is a challenge, because nonclinical results have to be delivered in time for the applications for the next clinical trials or for marketing approval. In some situations studies are conducted after approval, either as post-approval commitments or to support post-marketing safety evaluation. So nonclinical documentation support is usually needed throughout the compound's whole lifecycle.
Preclinical development planning is a complex task, and good preclinical planning will avoid potentially harmful and costly clinical trials.
How the programme depends on the molecule
Nonclinical studies are required for the development of all new pharmaceutical compounds, but their number and type depend on the compound's characteristics, in particular whether it is a small molecule or a biologic: under certain circumstances biologics are exempt from some pharmacokinetic and safety studies, although they often require additional case-by-case assessments. In general the need for nonclinical studies is assessed case by case, so a real programme is always tailored to the compound being developed.
The molecules can be divided into small molecules and biologics or large molecules, and the newer cell, gene and tissue therapies are each regulated as well.
For small molecules, early in silico predictions and in vitro and in vivo screens feed compound optimisation during discovery and also form the basis for managerial decisions to advance a molecule into GLP safety assessment; successful GLP studies enable the first human trials, the larger and longer phase II and III trials require more toxicological evidence including subchronic and chronic repeated-dose studies, and the studies needed for marketing approval such as carcinogenicity should be initiated well in advance of submission, usually during phase III. For biologics the evaluation requires fewer studies but may still become complex, because of limited experience with new molecules and modes of action or a lack of relevant species. Immunogenicity is another challenge, since it can lead to unwanted pharmacodynamic effects such as lack of efficacy, unwanted pharmacokinetic effects such as high variability and fast clearance, or adverse reactions, and immunogenicity testing can therefore constitute a large portion of a biologic's nonclinical programme.
In a set of small molecules one contract research writer looked at, a first-in-human package held a mean of 37 studies of pharmacology, ADME and toxicology together, although some companies sometimes have double that; in an example set of monoclonal antibodies there tend to be about half as many, because not much ADME or toxicology work is done on them.
Guidelines
The main guideline for the general setup of a nonclinical development programme is ICH M3, which describes the general principles, timing and standards for nonclinical safety evaluation. Because it is a fairly broad document, it is supplemented by a large set of specialised guidelines, the ICH S series, which detail the requirements for particular types of nonclinical study or specify programmes for certain compound classes or patient populations. For advanced therapy medicinal products there are several guidelines, and in May 2024 the overarching European guideline on them was still to come: a draft existed, with no indication of how long it would take to become final.
As of 2017 the ICH S series ran as follows. S1A, S1B and S1C cover the need for carcinogenicity studies, testing for carcinogenicity and dose selection for those studies, and were to be replaced by a single comprehensive S1 on rodent carcinogenicity testing. S2 describes the standard test battery for genotoxicity of small molecules and recommends individual tests. S3A covers toxicokinetics — the assessment of systemic exposure in toxicity studies, explicitly not nonclinical pharmacokinetics and metabolism testing — and S3B covers repeated-dose tissue distribution studies. S4 sets the minimum acceptable duration of chronic toxicity testing at six months for a rodent and nine months for a non-rodent. S5 is the key reproductive toxicity guideline and was under revision. S6 covers the safety evaluation of biotechnology-derived pharmaceuticals as a supplement to M3. S7A and S7B cover safety pharmacology and the potential for delayed ventricular repolarisation, and safety pharmacology studies may be treated as part of the toxicology programme and summarised in the toxicology written summary. Toxicity is evaluated according to S7A and S7B to predict short-term safety in people. S8 covers immunotoxicity testing for immunosuppression and immuno-enhancement, but not allergenicity or drug-specific autoimmunity. S9 gives the specific requirements and the toxicology study exemptions for anticancer compounds. S10 covers photosafety evaluation in detail, including the situations that require no experimental evaluation in biological systems. S11, then still to come, was to cover nonclinical safety testing supporting the development of paediatric medicines.
Nonclinical evaluation is further guided by cross-disciplinary documents. ICH M7, on the assessment and control of DNA-reactive mutagenic impurities to limit potential carcinogenic risk, complements the existing impurity guidelines ICH Q3A and Q3B with considerations for the risk assessment of potentially mutagenic impurities. The European guideline on strategies to identify and mitigate risks for first-in-human and early clinical trials supplies additional detail on the nonclinical programme intended to support entry into humans, including more precise requirements for assessing pharmacodynamics and for determining the starting dose. Many nonclinical pharmacokinetic in vitro evaluations are mandated by drug–drug interaction guidelines. The European agency has a dedicated page listing its nonclinical guidelines comprehensively.
Although these guidelines are extensive and appear comprehensive, they cover only a subset of the nonclinical studies that appear in a submission package and give no general account of how the nonclinical programme fits into the whole development programme.
Good laboratory practice
Preclinical studies have to comply with good laboratory practice so that their results are reliable. Nonclinical studies are performed, often simultaneously with the chemistry, manufacturing and controls work, following specific protocols and employing rigorous control processes that must be documented for the results to be accepted.
Non-pivotal studies include pharmacokinetic studies for species selection, dose range-finding studies for the in vivo toxicology work, and screening genotoxicity studies; these are mostly not conducted under good laboratory practice. Pivotal studies are mainly the regulatory in vivo toxicology and genotoxicity studies, together with the pharmacokinetic work that is part of the toxicology studies, and these have to be performed with validated methods under GLP.
The OECD GLP guideline, the nonclinical sister to good manufacturing and good clinical practice, is a common reference point for good laboratory practice, but regional legal implementations can vary significantly. Interpretation of the GLP guidelines can differ between countries, and a study conducted in a country outside the OECD mutual acceptance of data system may not be accepted by some health authorities: in 2017 South Korea and the United Kingdom did not accept GLP studies conducted in China. Because the methodology of nonclinical studies overlaps significantly with industrial toxicity testing, regulators also expect OECD guidelines to be considered where applicable, which include not only the GLP guideline but also many internationally agreed and validated toxicity testing methods.
Study types
Before a compound reaches humans the package generally contains general toxicity studies, genotoxicity, receptor interaction work, pharmacokinetics, protein binding, metabolic stability data and safety pharmacology; later it is extended by further studies, carcinogenicity among them.
Pharmacodynamics asks what the drug does to the body, and pharmacokinetics what the body does to the drug; both are among the initial pharmacology studies. Pharmacodynamics is also defined as the relationship between the concentration of a drug in the body and its biological effect, which is the dose response. It covers how potent and how efficacious the drug is with regard to its desired pharmacological effect, including safety aspects and adverse effects, and so it establishes the therapeutic index — the ratio of the dose causing toxicity to the dose eliciting a therapeutic effect. Ideally that index is large, indicating a wide therapeutic window.
A drug's effect is determined by the amount of active drug present in the body and particularly at the target site, which depends on absorption, distribution, metabolism and excretion, and pharmacokinetics describes the resulting change in plasma concentration over time; ADME profiling is critical for establishing the dose range and administration schedule for the clinical phases that follow. Most drugs are given orally and must be absorbed in the gastrointestinal tract to enter the bloodstream, and on the way to the target they reach the liver, where first-pass metabolism reduces the concentration and therefore the bioavailability before the drug enters systemic circulation; intravenous administration bypasses that effect and gives greater bioavailability. Distribution through the body is determined by the drug's affinity for plasma proteins, its molecular properties and polarity, and tissue vascularisation. Metabolism is the chemical alteration of the parent drug into pharmacologically active or inert metabolites. Information on clearance is critical to ensure adequate long-term dosing and appropriate steady-state concentrations; clearance is achieved mainly by the renal and hepatic routes, although pulmonary clearance plays a major role for volatile drugs such as anaesthetics, and concomitant disease, lifestyle factors and a patient's age can all affect it and are frequently studied in later clinical stages. Steady state is reached when the rate of clearance equals the rate of absorption; maintaining a stable steady state is typically desirable and can be achieved through repeated dosing. The drug and its metabolites are excreted mainly in urine or faeces. For pharmacokinetics, the area under the curve and Cmax are considered the most important parameters.
ADME studies establish how the drug is absorbed, distributed, metabolised and excreted. Some metabolism data are worth having before any animal experimentation starts: how stable the compound is and which metabolites can form, which can be studied in vitro for several species including human. That matters because the species used for in vivo toxicology should cover all the metabolites that humans form, since not only the parent compound but also its metabolites can be toxic. Reaction phenotyping can already identify the enzymes involved in the metabolism, which gives early information on the drug–drug interactions that may be encountered in the clinic. Examples of what is tested include which enzymes interact with the drug, the drug's ability to bind protein in cells, and the metabolites formed as it is broken down, alongside the basic routes of absorption, distribution, metabolism and excretion.
Safety pharmacology investigates potential undesirable pharmacodynamic effects on vital body functions — cardiovascular, respiratory and central nervous system. These are mostly single-dose studies, and their results are evaluated through an integrated risk assessment across all the nonclinical study results. Where an effect appears, follow-up supplemental studies are performed and will be in the package.
Toxicology asks whether a potent compound is safe. Preclinical toxicology studies identify the treatment regimen associated with the least degree of toxicity and so determine a suitable and safe starting dose for clinical trials, and they can also establish biomarkers for monitoring potential adverse effects later. Examples of toxicity studies include testing for adverse effects on specific organs or organ systems, carcinogenicity, and effects on reproduction.
Genotoxicity establishes whether the compound binds directly to DNA or acts indirectly on processes such as DNA replication, and it matters because a genotoxic substance may be carcinogenic. Particularly in smaller companies a genotoxic signal is often considered a showstopper for the development of the compound; a new painkiller, for example, should not be genotoxic. The package contains both in vitro and in vivo studies. For biologicals, genotoxicity testing often does not need to be performed. The genotoxic effect is usually studied through the compound's potential to induce mutations in yeast-based in vitro systems.
Reproductive toxicology is mostly three types of study, which between them cover the entire reproductive cycle. The cycle studied runs from before mating through to weaning and the sexual maturity of the pups. The studies are performed so that women of childbearing potential can be included in clinical studies, and there is a large table of when that inclusion becomes possible. Reproductive toxicity studies typically involve rats.
Carcinogenicity studies are required where the expected clinical use is continuous for at least six months, or intermittent to a total exposure of at least six months. They are very large and very expensive — daily dosing for two years in large groups of animals — so they are mostly conducted only towards the end of the development period. By 2024 some alternatives had become possible: running only one 24-month study together with one shorter one, or taking a weight-of-evidence approach, which the guidelines allow but the regulators do not always accept.
As of 2017, a carcinogenicity assessment typically consisted of a two-year rat study and a six-month transgenic mouse study, each preceded by dose-range-finding work. Carcinogenicity studies typically involve rats and, because a tumorigenic effect may only become evident after prolonged exposure, comprise continuous administration for a minimum of six months.
Later in development, specialised areas are added according to what the compound does. A drug with phototoxic potential needs appropriate photosafety studies, usually in guinea pigs. A drug that crosses the blood–brain barrier into the central nervous system has the potential for addiction, and there is a whole array of animal studies for that. For a drug going into a paediatric population, studies in juvenile animals are nowadays sometimes needed.
One experienced writer has seen development delayed many times because impurities had not been properly covered and discussed, and new genotoxicity and toxicology studies with the impurities are sometimes needed.
Species
Regulators require safety testing in at least two mammalian species, one rodent, such as a mouse, guinea pig or hamster, and one non-rodent, before a first-in-human trial can be approved. In 2022 regulatory agencies recommended testing in two animal species to guarantee safety and efficacy, initially a rodent and then a larger non-rodent mammal; most laboratories relied on mice, and dogs were usually the second species, since a dog's metabolism and response to drugs are closer to humans'. For small molecules the general toxicity studies are run in two species, one rodent and one non-rodent, while for biologics and advanced therapies one relevant species is mostly sufficient.
The programme as a whole can involve more than two species, because particular questions need particular animals: studies have been run in woodchucks, one of the few suitable models for hepatitis B. For biologics the only relevant species may be a non-rodent, usually a non-human primate, and the rodent requirement can then be waived. Health authorities may also ask for the non-rodent studies to be supplemented with data from transgenic mouse models.
Species selection is always a big problem in nonclinical development, or a matter of debate. The species should be relevant to the human situation, the route of administration should in general be the one that will be used in the clinic, and sufficient background information on healthy animals is needed. The principle of animal welfare, the three Rs of reducing, refining and replacing animals where possible, is becoming more and more important. There is a tendency to reduce animal numbers, and whether the authorities will really accept that is a matter of debate.
Species are also chosen according to the intended route of administration and the anticipated absorption, distribution, metabolism and excretion results, the goal being a species with a specific characteristic that will closely approximate the outcome in a particular human organ. Animal testing is closely monitored by animal care boards, and the minimum number of animals required to achieve the study goals is used.
Practical constraints then narrow the field: a lion is too dangerous, oral dosing is difficult in animals that refuse to swallow, and dosing on a milligram-per-kilogram basis rules out very heavy animals because of the quantity of drug substance the chemistry and manufacturing colleagues would have to supply. What is left in practice is mice and rats, dogs, rabbits, and pigs for applications on the skin. By 2024 dogs were becoming more and more of a problem, particularly in the United Kingdom, and were often replaced by minipigs. Biologics often end in monkeys, because a biologic is made to be active in humans and the species has to be a relevant one, where there is some pharmacodynamic activity.
General toxicity study design
The design of a general toxicity study depends on the type of drug, a small molecule and a biologic being set up slightly differently, and it matters for what is tabulated later.
In vivo toxicology work starts with single-dose studies, to identify the organs that might be subject to drug toxicity, and continues with repeated-dose work whose regimen ideally mimics the intended clinical design in treatment duration, schedule and route of administration.
ICH M3(R2) recommends an extended single-dose toxicity study, which besides acute toxicity determines clinical, chemical, haematological, haemostatic, toxic, kinetic and other parameters, gives a wider overview than the common approach, and produces results compatible with those of repeat-dose studies. In most cases single-dose toxicity can be evaluated in escalation-dose or short-dose experiments. The OECD methods do not always require a 50% lethal dose to be determined.
A study mostly runs three dose levels plus a control group. The low dose should be sufficient to produce a pharmacodynamic effect. The high dose can be selected in various ways, and some toxicity should appear there, because that guides which parameters have to be studied in the clinical study. The mid dose is mostly the geometric mean of the two.
The parameters investigated are mortality, morbidity, body weight, food consumption, clinical signs, haematology and clotting, and clinical chemistry, and there is always a toxicokinetic evaluation to relate effect to exposure. At the end the animals are examined across about 40 to 50 tissues and organs for any effect of the compound on them.
Interpreting toxicology findings
A toxicology report contains a great deal of detail and many findings, not all of them drug-related. When it comes to interpretation there is a whole array of potential findings, to be described in an ordered way: real toxic effects relevant to human safety; effects specific to the rat or the dog and not relevant to humans; pharmacological effects; biochemical effects; mechanistic effects; effects that cannot be explained; and high-dose effects. With biologics, effects on the immunological system can be expected. An adverse effect related to the compound's pharmacological action may be less of a problem than a real toxicological adverse effect.
Several specific points need care. A cardiovascular, hepatic or renal finding needs the relevance of the finding emphasised, and an equivocal result needs discussing properly. A statistical change does not necessarily mean a real finding; it can be natural variation. Whether a finding appears in all the groups that got the drug or only in the top groups needs writing carefully. Organ weights need to be stated as absolute, relative or both, because an absolute change may only reflect the animal losing body weight. And findings have to be discussed across species and studies.
In the reproductive studies, fetal findings may be purely due to maternal toxicity rather than real teratogenic effects on the skeleton and viscera. Many small findings, such as the additional ribs often seen in rabbits, are variations that are not biologically relevant, and they need discussing against real malformations. What eventually goes into the product label is weighed by looking across all the data, including whether findings appear in both species, rat and rabbit. In the carcinogenicity studies, the dose levels, especially the high dose, need discussing, and so does survival: rats dosed for two years are very old by the end, and there have to be enough survivors to assess for tumours. If few tumours appear, the question is whether the result is valid; if tumours appear, how relevant they are, whether they are real and whether they have significance for human safety. Rodents sometimes develop tumours at sites that do not occur in humans, such as the forestomach, which rats have and humans do not. Positive genotoxicity findings need proper interpretation: whether they are relevant, whether similar compounds produce them, and how they relate to the carcinogenicity findings, ending in a weight-of-evidence approach. Tyrosine kinase inhibitors in oncology are always positive, and that is acceptable because the drugs go into cancer patients.
The first dose in humans
Translating the package into a human dose is itself governed by guidelines from both the European and United States agencies, and there are two main approaches.
The simpler one starts from the no observed adverse effect level — the highest dose in the toxicology study at which no adverse effects were seen. That is converted by body surface area, using a table taken from the guideline, into a human equivalent dose, and a safety margin is then applied to give the starting dose for the first-in-human trial. In outline, various dose levels are tested to determine the no-observed-adverse-effect level and the minimum lethal dose, which are later used to calculate an acceptable dose for humans.
By 2024 more and more was done by the other approach, MABEL, the minimum anticipated biological effect level. All the animal and in vitro pharmacology data are gathered, together with pharmacokinetic data in multiple animal species and anything else relevant such as plasma protein binding; an extrapolation is made from animal to human; a dose–response model is built; and other factors such as species differences and a safety margin for uncertainties are put in, producing a human dose relationship. The modelling is an art in itself. A safety margin, mostly a small one, is then applied to the MABEL value to set the first dose level in the clinical studies. The proposed clinical exposure then has to be compared with what was found in the toxicology studies, to make sure that no toxicological effects are seen at the dose being started from.
The European guideline on first-in-human clinical trials sets out principles for calculating the starting dose in humans, for the subsequent dose escalation and for the criteria for the maximum dose; they stood both in the draft revision the European Medicines Agency proposed in November 2016 and in the revision it published in July 2017. On the nonclinical side the draft proposed better integration of pharmacokinetic, pharmacodynamic and toxicological data into the overall risk assessment, and set out the role of nonclinical data in defining the estimated therapeutic dose, the maximal dose, and the dose steps and intervals.
Nonclinical data during clinical development
Experienced nonclinical writers hold that nonclinical data do not face less scrutiny because some clinical data in healthy volunteers already exist: the outcome of the nonclinical studies determines quite a few of the factors that have to be studied clinically, and a regulator has to assess whether any signals from the nonclinical data were indeed studied in the clinical work, so ignoring them is not an option. In their view, unless the clinical work has been very elaborate and has focused on all the endpoints that might be a problem, regulators will not accept it alone and will want to look at the nonclinical data as well. The FDA can look again at the nonclinical data while a phase 3 study is under way. In one study in rare disorders and endocrinology, it found eye changes in the animal data and requested that the protocol be revised so that all patients had eye investigations, and the company had to commission eye clinics to run them; a participant in the study called this very important for the clinical development that followed. In an older case, a question about QT prolongation during phase 3 sent a team back through all their animal ECG data to check for a missed signal; they had not missed one.