A toxicology program has one job: to show that a drug is safe enough for a specific use in specific people. That makes the first clinical trial the starting point for every study that comes before it, and it is why the route, the dosing interval and the patient population need to be pinned down, or written down as assumptions, before the first study is designed.
A biotech company can have strong efficacy data and still not be ready to start toxicology. The usual reason is that nobody has pinned down how the drug will be used in people. The route might still change. The dosing interval is a guess. The first trial may include patients whose risks nobody has thought through yet.
Each of those parameters decides what the animal studies have to show, because a toxicology program has one job: to show that the drug is safe enough for a specific use in specific people. If the use is still vague, the studies end up answering a question the company has not finished asking. And toxicology studies are expensive questions to ask twice.
So when a sponsor reaches out, I start with the clinical expectations and work back to the studies. That is my position throughout this article, and it has a limit I will come back to: you rarely know everything about the trial during the planning phase, so the plan has to say which parts are known and which are assumed. Below, I walk through the nine steps to go from the first description of the trial to a plan the whole team can act on.
The nine steps at a glance
- Describe the clinical trial before you plan the tox studies
- The first assumption to test is exposure
- An exposure number only means something in a relevant species
- With the relevant de-risking and the species settled, put the studies in order
- How far the sequence runs depends on the trial and the patients
- What the studies must deliver: a starting dose and a monitoring plan
- All of it rests on the material and the methods
- Take the biggest remaining uncertainties to the regulator
- Keep one plan, and keep it current
Describe the clinical trial before you plan the tox studies
The first step produces a description of the trial, and it does not need to be a finished clinical protocol. It does need to be clear enough to design from. Take a company planning a tablet taken once a day for several weeks. That plan raises different safety questions than a monthly infusion would, or a single dose that is meant to keep working for years.
To get that description, I work through a starting brief: the indication, how the drug works, who will be in the first trial, the route and delivery method, the dose range, the dosing interval, how long treatment lasts, and how long patients are followed afterwards. Together, the answers make up what I will call the clinical picture, and every later decision in the program refers back to it. For each item I ask the same thing. Is this based on data, or is it a working assumption? A once-daily schedule that comes out of an animal PK study deserves more confidence than one chosen because once a day is convenient.
The brief also has to say what the first trial is meant to learn, because “Phase 1” is only a label. A single ascending dose study, a multiple ascending dose study and a protocol that combines the two can each need different supporting evidence. The regulators frame it the same way. ICH M3(R2) ties nonclinical testing to the scope and length of the human trial, which works best for conventional drugs; biologics also follow the product-specific principles in ICH S6(R1), and cell and gene therapies have their own guidance.1–3
The first assumption to test is exposure
The assumption that most often turns out to be untested is the dosing schedule, and the way to test it is to measure exposure. Dose is what you give a patient. Exposure is what the body actually sees: how much drug reaches the blood and how long it stays there. Safety follows exposure, so a schedule that has never been checked against exposure is a guess.
Checking it starts with early pharmacokinetic (PK) work, which tells you how well the drug is absorbed, how long it lasts and what the concentration curve looks like after a dose. Toxicokinetics (TK) is the same kind of measurement taken inside the toxicity studies, so that a finding can be linked to a drug level. ICH S3A gives TK exactly this role: interpreting toxicity, and judging whether animal findings are relevant to humans. TK provides measures such as the peak concentration (Cmax) and the total exposure over time (AUC), and these serve as the basis for setting safety margins for humans.4
Going back to the once-daily tablet: if you are expecting to dose daily, but the drug stays in the body for a week, there is no point in dosing daily. This is a point I often emphasize. How long the drug lingers is only part of the answer, though. You also need to know how long it stays on its target, how long the effect lasts, what concentration it needs to work, and whether it accumulates.
Accumulation matters a great deal for safety. If concentrations climb with every dose, the fact that animals tolerated the first dose tells you little about the tenth. So for any drug given repeatedly, it is important to determine whether exposure rises in step with the dose, and whether the early studies run long enough to show what happens over time. That is why it is critical to evaluate PK early in the program, in short-term repeat-dose studies.
The same questions follow the drug into the clinic. Decide in advance what you will do if exposure in people comes back higher than expected, what you need to see before raising the dose, and whether your blood sampling times will capture the full profile. EMA guidance treats the starting dose, dose escalation and maximum exposure as connected parts of managing risk in a first-in-human trial, and that is the right way to think about it.5 What you should have at the end of this step is a written exposure hypothesis: what you expect, the data behind it, and how you will test the rest.
An exposure number only means something in a relevant species
Testing that hypothesis means measuring exposure in animals, which raises the next question: which animals? A species you have used before is not automatically informative for a new molecule. If the drug does not do in that species what it does in humans, a clean study there tells you very little.
For biologics, relevance comes down to two things: whether the drug binds its target in that species, and whether binding has the same effect there as in humans. ICH S6(R1) discourages testing in species where the drug is not active, and it allows justified alternatives when the clinical candidate is active in only one species.2
A well-documented case shows both why that comparison matters and where it reaches its limits. TGN1412, a CD28-targeting antibody later renamed theralizumab, caused a severe, body-wide inflammatory reaction in six healthy volunteers during its first clinical trial. Later research found that CD28 is expressed differently on effector memory T cells in humans than in the nonhuman primates used for testing, which offers an explanation for why the animal studies did not predict the reaction.6,7 That difference could not reasonably have been foreseen with the methods of the time, so the case is a reminder of what animal studies can miss rather than an example of poor planning.
Small molecules need the same check from a different angle: does the proposed species handle the drug, through metabolism and exposure, in a way that tells us something useful about humans? Efficacy models need their own reasoning again. A model that skips the biological step your drug acts on will show no effect, even if the drug works. And a model that showed efficacy is not, for that reason alone, the right species for safety.
With the relevant de-risking and the species settled, put the studies in order
Once you know what you need to measure and where you can measure it, the remaining decision is the order of the studies. The early ones exist so that the large, expensive ones can be designed well. In my experience that usually means some combination of in vitro work, PK, exploratory tolerability and short repeat-dose studies before committing to a big study.
Each of those steps should answer a question that shapes the next one. Can we reach the exposure we intend? Does the formulation work? What changes when the drug is given repeatedly? Which observations need a closer look? By the end, you should know enough to justify the dose range and the sampling schedule of the definitive study.
I have seen what happens when a step is left out. Sponsors have moved from a short study straight into chronic studies to save time. The doses turned out to be wrong, there were findings even at the lowest dose, and the chronic study could not support the decision it was run for. The shortcut created more work.
That does not mean there is a fixed ladder every drug has to climb. The sequence depends on the molecule, the data you already have, the clinical picture and the guidance that applies. M3(R2) guidance itself is flexible here: the maximum tolerated dose is not the only acceptable basis for choosing the high dose, and limits based on exposure or on what is feasible to administer can be appropriate too.1
So what does this mean for your own program? Agree in advance what evidence lets you move on. Before each study, I want the team to say which uncertainties it has to resolve and which results would mean changing course. A study can be run perfectly and still leave the program stuck if nobody asked those questions first. For a startup, there is a second benefit: spending is tied to evidence milestones, and you can explain why a smaller study comes before a larger commitment.
How far the sequence runs depends on the trial and the patients
How many of those steps you need before the first trial, and how long the longest study has to be, comes straight from the clinical picture. The first trial is one milestone in a longer safety program. The useful move is to separate what that trial needs from what you will need later for longer treatment, a wider patient group, or a marketing application.
Treatment length sets the first boundary. For conventional drugs, M3(R2) guidance generally links the length of the repeat-dose toxicity studies to the length of clinical dosing; two-week studies, for example, generally support trials of up to two weeks, with some regional alternatives and exceptions. Longer clinical dosing needs correspondingly longer studies. Do not assume the same rule carries over to a biologic or a gene therapy.1
The patients set the second boundary. The risk that is acceptable in healthy volunteers is different from the risk that is acceptable in patients with advanced cancer and few options, which is why ICH S9 gives that setting a separate framework. Its flexibility does not extend to every oncology program.8 Children are a different case again. There, the question is whether the existing evidence covers the risks to organs that are still developing, and ICH S11 answers it by weighing the youngest intended age, the pharmacology, the length of treatment and the data already available. Juvenile animal studies are not automatically required.9
And some products stretch the idea of “treatment length” altogether. For cell and gene therapies, a single dose can call for long observation because the biological activity continues. The plan may need to cover where the product goes in the body, how long it persists and how the transgene is expressed. FDA’s preclinical guidance discusses these points in relation to the particular product and the proposed trial.3
My recommendation is to list the later requirements at the start and then schedule each one according to the decision it supports. That way you move efficiently toward the first trial and still keep the next stage in view.
What the studies must deliver: a starting dose and a monitoring plan
Everything up to this point serves two outputs for that first trial. One is a first human dose you can defend. The other is a clear instruction to the clinical team about what to monitor, when, and how new findings will affect further dosing.
For the dose, the best-known method is in FDA’s starting-dose guidance for trials in healthy adult volunteers. You take the no observed adverse effect level (NOAEL) from the animal studies, which is the highest dose that caused no adverse effects. You convert it to a human equivalent dose, and then you apply a safety factor. Typically a safety factor of 10 is applied, although it can vary with the population: it can be lower for patients with advanced cancer, and a larger margin may be applied when the trial includes children.10
That method starts from toxicity, and for some drugs toxicity is the wrong place to start. They are biologically active at exposures well below those that harm animals. For these, EMA guidance discusses approaches based on the minimal anticipated biological effect level (MABEL), the pharmacologically active dose, and how uncertain the relevance to humans is.5
Monitoring comes from the same studies, which is why I want the toxicology and clinical teams to talk before the designs are locked. If a finding could change who is eligible, how long subjects are observed, which lab tests are run, or how the dose is escalated, we should ask whether the study as designed will give enough information to interpret it. Safety pharmacology belongs in that conversation too. It covers the vital functions, in particular the heart and circulation, breathing, and the central nervous system, and ICH S7A allows these assessments to be built into other studies where that is scientifically justified.11
It is tempting, at this point, to measure everything. My design principle is the opposite: a good tox program has all the endpoints that are needed, but not too many. Every relevant safety question has to be covered. Extra exploratory measurements need a clear purpose and a practical way to collect them, because every added procedure can make the results harder to interpret. The test is the link back to the clinical protocol. Can the team explain how the animal findings support its monitoring and escalation decisions, and where that support runs out?
All of it rests on the material and the methods
A starting dose and a monitoring plan are only as good as two things that sit underneath every study: the material that was tested, and the methods used to measure it. A well-designed study loses much of its value if you are unsure what went into the animals.
I have seen that uncertainty first-hand. In one early program, we checked the candidate molecules before going ahead and found that their identity and purity were not what we expected. The chemistry needed more work. Taking that step back kept us from building later experiments on a foundation we could not trust.
That experience does not mean every early experiment needs material from the final clinical manufacturing process. It means the sponsor should understand what its material represents, and how a change could affect the relevance of the results. Manufacturing, formulation, vehicle and delivery choices should be discussed with the toxicology team before important studies are committed.
Delivery deserves a special mention, because it is easy to treat as a detail. I worked on a powder-formulation program that reached phase 3, where a change to the drug delivery device led to a request for another clinical study. The formulation itself had not changed. Just changing one minor thing during the program can have a big impact, because it changes the questions your existing evidence has to answer.
The measurements have their own dependencies. ICH M10 covers validation and sample analysis for the methods used to quantify a drug. Sensitivity, selectivity, interference from the sample itself, and stability all need to be considered for the samples and the concentration ranges the study will produce.12 In practice, I ask who owns each method, where it will run, whether it has to be transferred to another lab, and how the samples must be handled. A delay in method development, or a shipping plan that does not work in practice, can hold up the whole schedule.
Take the biggest remaining uncertainties to the regulator
By now the plan is nearly complete, and it will still contain a few assumptions that data alone cannot settle. Those are what an early regulatory meeting is for. It is most useful when you bring a specific proposal and can say exactly which uncertainty you want resolved. Asking in general terms which studies to run usually gets you less.
The preparation follows from the steps above. I recommend a concise account of the intended clinical trial, the relevant pharmacology and exposure data, the proposed species and study designs, and the reasons for any departure from the usual approach. Then aim the questions at the places where the agency’s answer could change a significant investment. Does the proposed model adequately address the mechanism? Is the study duration appropriate for the intended clinical exposure? Does the package cover the planned patient population? What more is needed to support the proposed route or delivery approach?
FDA’s Office of Therapeutic Products makes the same link in its pre-IND resources. Its suggested questions cover species and models, the design of the definitive toxicology study, the adequacy of the overall preclinical program, and the clinical starting dose and escalation plan. Those resources apply to the products that office handles, but the same preparation is useful more broadly.13
Running a tox study just for the purpose of running a tox study, and then getting to the agency with the wrong species or the wrong dosing regimen, is a waste of time, money and animals. A timely discussion helps prevent that. It comes with one condition, though. The advice applies to the plan you described. It does not guarantee that future results will be accepted, and it does not cover material changes to the product or the protocol.
Keep one plan, and keep it current
That condition is the reason the last step is about upkeep. Advice, study designs and dose calculations all refer back to the clinical picture, so when the picture changes, someone has to ask what else changes with it. I recommend keeping a record of the changes that matter. A new route, dosing interval, formulation, patient population or expected exposure should trigger a review of the supporting evidence. The team can then decide whether the existing work still applies, whether more investigation is needed, or whether to go back to the regulator.
A record like that only works if the plan has named owners. The sponsor, the toxicologist, the clinical lead, the regulatory adviser, the CMC team and the CRO should each know what they are responsible for and what information they owe one another. The CRO’s study calendar is only one part of the schedule. Material availability, method readiness, reporting, interpretation and submission preparation have to be coordinated as well.
Here is the test I would apply before a startup approves substantial toxicology work. Can the leadership team explain the proposed first trial, the evidence behind its assumptions, and the uncertainties that remain? If so, the early work can stay proportionate to the decisions it supports, and the larger commitments can follow once the evidence makes their design credible.
I said at the start that this approach has a limit: you never know everything about the trial during the planning phase. The nine steps do not remove that limit. They make it visible. The route, the dosing interval and the patients, the three open points I began with, are each supported by data or written down as an assumption with a plan to test it.
Do not cut corners. Even if it takes longer, do things well, and do them once, so you do not hit a wall and have to start again from scratch. For a company preparing its first clinical trial, that means describing the trial early, testing the assumptions that matter, and designing each study to support the next decision.
Key takeaways
- Start from the trial. The route, the dosing interval, the length of treatment and the patients decide what the studies have to show.
- Separate data from assumptions. For each item in the clinical picture, mark whether it rests on data or is a working assumption.
- Test exposure first, in a relevant species. A dosing schedule that has never been checked against exposure is a guess.
- Put the studies in an order that informs the next decision. Agree before each study what evidence lets the program move on.
- Confirm the material and the methods, then take the remaining uncertainties to the regulator. Bring a specific proposal, and review the plan whenever the clinical picture changes.
Talk to us about your program
At KreaMedica, this is where early program discussions are most valuable: bringing the clinical picture, the nonclinical evidence and the practical dependencies into one coherent plan before the sponsor commits to major studies.
KreaMedica is an integrated development partner for lean biotechnology companies. We help seed- through Series A-stage companies plan and execute complex drug-development programs without building a large internal development organization — bringing deep expertise across toxicology, preclinical development and CMC, a curated network of CROs, CMOs and specialist consultants, and hands-on program leadership. We execute, not just advise.
About the Author
Dominic Poulin (Senior Preclinical Toxicologist, KreaMedica)
Dominic is KreaMedica’s Senior Preclinical Toxicologist, with over 25 years of experience in the preclinical CRO industry. His career spans hands-on laboratory work and study direction, with expertise in inhalation toxicology, other major routes of administration, and in vivo pharmacology models. He brings practical troubleshooting experience and a strong understanding of CRO operations and regulatory expectations to help clients design and execute robust preclinical studies.
References
1. International Council for Harmonisation. M3(R2) Nonclinical Safety Studies for the Conduct of Human Clinical Trials and Marketing Authorization for Pharmaceuticals. 2009; FDA implementation, January 2010. Source
2. International Council for Harmonisation. S6(R1) Preclinical Safety Evaluation of Biotechnology-Derived Pharmaceuticals. 2011. FDA hosts the original S6 guideline and its 2012 addendum implementation. Source
3. US Food and Drug Administration. Preclinical Assessment of Investigational Cellular and Gene Therapy Products. Guidance for Industry. November 2013. Source
4. International Council for Harmonisation. S3A Toxicokinetics The Assessment of Systemic Exposure in Toxicity Studies. 1994; FDA implementation, March 1995. Source
5. European Medicines Agency. Guideline on Strategies to Identify and Mitigate Risks for First-in-Human and Early Clinical Trials with Investigational Medicinal Products. EMEA/CHMP/SWP/28367/07 Rev. 1. 2017; effective February 2018. Source
6. Suntharalingam G, Perry MR, Ward S, et al. Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412. New England Journal of Medicine. 2006;355:1018–1028. doi:10.1056/NEJMoa063842. Source
7. Eastwood D, Findlay L, Poole S, et al. Monoclonal antibody TGN1412 trial failure explained by species differences in CD28 expression on CD4+ effector memory T-cells. British Journal of Pharmacology. 2010;161(3):512–526. doi:10.1111/j.1476-5381.2010.00922.x. Source
8. International Council for Harmonisation. S9 Nonclinical Evaluation for Anticancer Pharmaceuticals. 2009; FDA implementation, March 2010. Source
9. International Council for Harmonisation. S11 Nonclinical Safety Testing in Support of Development of Pediatric Pharmaceuticals. 2020; FDA implementation, May 2021. Source
10. US Food and Drug Administration. Estimating the Maximum Safe Starting Dose in Initial Clinical Trials for Therapeutics in Adult Healthy Volunteers. Guidance for Industry. July 2005. Source
11. International Council for Harmonisation. S7A Safety Pharmacology Studies for Human Pharmaceuticals. 2000; FDA implementation, July 2001. Source
12. International Council for Harmonisation. M10 Bioanalytical Method Validation and Study Sample Analysis. 2022; FDA implementation, November 2022. Source
13. US Food and Drug Administration. OTP Pre-IND Meetings. Source