From preclinical toxicology to clinical safety: The path for oncology IND submissions
Oncology drug development programmes are built for speed because the patients they ultimately serve have few remaining options. Compressed timelines are a necessity, and they shape the way non-clinical evidence is generated, interpreted, and translated into first-in-human (FIH) decision-making. Toxicology must therefore be calibrated to two realities: first, that oncology agents are often intrinsically toxic at effective exposures; and second, that early clinical testing occurs in patients with advanced disease, not in healthy volunteers.
The ‘toxicology bridge’ that leads from preclinical data to a defensible FIH dose always focuses on practical choices that balance pace and safety. The regulatory context surrounding these choices is intentionally different for oncology drugs than for non-oncology drugs. Successfully navigating the process for drugs with a higher risk-benefit ratio requires a tailored non-clinical programme design that is intentionally minimal, targeted, well-justified, and anchored to the intended clinical plan.
Minimum preclinical toxicology requirements of ICH S9
For cancer drugs, the ICH S9 guideline allows a streamlined set of preclinical studies to support an IND (the application to begin human testing). The cornerstone is a 4-week repeat-dose toxicology study run under Good Laboratory Practice (GLP) in one or two pharmacologically relevant species. The pivotal studies should embed basic safety-pharmacology observations within routine in-life and pathology assessments and, critically, incorporate toxicokinetics (TK) sufficient to define exposure margins at the intended clinical schedule.
In addition, one disease-relevant tumour model with a clear biomarker readout demonstrates efficacy at plausible exposures. Extra tests are limited: hERG testing for heart-rhythm risk is only run for compounds with a structural alert or early screenings that raise concern, and genotoxicity testing is only needed for non-cytotoxic drugs, since direct DNA damage is inherent to many cytotoxic mechanisms.
What does this lean package achieve? The GLP study identifies target organs, characterises the pattern and severity of toxicity over the intended dosing cadence, and surfaces the practical monitoring parameters that should appear in the clinical protocol. TK closes the loop between dose, exposure, and toxicity, enabling safety margin calculations that justify a starting dose yet leave room for escalation.
Finally, it is essential to recognise the scope limits of S9. These guidelines apply to anticancer treatments for patients with advanced disease and do not cover those for prevention, symptomatic treatments, or vaccines. Most gene therapies fall outside the guideline as well, with the notable exception of CAR-T cell therapies, which are relevant within the oncology paradigm, but draw on additional expectations discussed below.
Study design adjustments based on modality and mechanism
The overall approach is consistent across modalities, but test species and endpoint details will vary. Small molecule drugs are typically tested in rodents and dogs, whereas large molecules like monoclonal antibodies, many antibody-drug conjugates, and nucleic acid therapeutics often require a more human-relevant species. For these larger-molecule drugs, studies should prospectively incorporate immunogenicity assessment and an anti-drug antibody (ADA) strategy, allowing for the correct interpretation of changes in exposure and side-effect profiles.
Agents that stimulate the immune system require additional forethought. For immune agonists and T-cell engagers, toxicity related to immune activation should be anticipated and planned for, rather than treated as an outlier. Non-clinical designs should embed immune-related endpoints like cytokine panels, clinical observations consistent with cytokine-mediated reactions, and predefined rescue plans. Their inclusion will help distinguish manageable, mechanism-linked effects from signals that warrant dose holds or discontinuation.
Cell therapies occupy a slightly different lane. CAR-T products generally do not require traditional in vivo repeat-dose toxicology unless the antigen receptor is novel and brings unique uncertainty. For this type of therapy, non-clinical work must emphasise in vitro T-cell activation and cytokine-release assays, used to estimate MABEL when appropriate, alongside tumourigenicity assessments, which are addressed under ICH S12, rather than S9. The common thread is a goal of proportionality, studying only what is necessary to achieve a safe and informed entry for patients.
ICH S9 guidance on duration and recovery
In oncology, the dose schedule, in addition to dose duration, drives nonclinical design. The enabling study should mirror the intended clinical cadence closely enough that exposure-toxicity relationships can be applied to the first trial. For instance, if the clinical plan requires daily dosing for five days every three weeks, the enabling study should administer five consecutive daily doses.
Recovery groups, by contrast, are not universally required for oncology drugs. Complete histologic or functional reversibility is not a prerequisite for moving the process forward. Recovery cohorts should be reserved for situations in which severe toxicity occurs near expected doses, and recovery cannot be reasonably expected based on mechanism, organ regenerative capacity, or historical experience.
Finally, the single-species approach under S9 is not a shortcut, but rather a deliberate design principle aimed at getting drugs to reach patients faster while focusing on the most informative model. A second species may be added when scientifically necessary. For example, it may be appropriate to extend studies to an additional species for drugs like radiopharmaceuticals, in situations when data from a more human-relevant species is needed, when dedicated assessments of a human-unique major metabolite is necessary, or when specific toxicities warrant additional evaluation.
Determining clinical starting dose and safety margins
In any first-in-human study, the aim is to choose a starting dose that is acceptably safe, yet still informative, then escalate carefully. Oncology differs in starting point, data source, and risk tolerance, given the advanced cancer diagnosis of study participants. In practice, over 90% of oncology programmes use HNSTD/NOAEL rules from in vivo data, but NOAEL is often secondary or absent as the goal is not to find a no-effect level, but to identify a safe, meaningful starting point. Studies for biologics typically set the first dose at about one-sixth HNSTD, and for small molecules teams often start at about one-tenth of the dose that caused severe toxicity in ten percent of in vivo subjects. In all cases, toxicokinetic data must be monitored to confirm that the planned human exposure remains within an acceptable safety margin.
MABEL (minimal anticipated biological effect level), another potential starting point, has a defined but narrower role. It is not the default in oncology, but it is mandated when there is a high risk of excessive pharmacological reactions such as cytokine storm, severe on-target damage to vital tissues, or irreversible immune activation. In such cases, in vitro study data on potency, receptor occupancy, and cytokine-release inform a lower, biologically anchored starting dose, with careful, pre-specified increments.
Warning signs that a dose is too high are well known. Ongoing weight loss, low activity or a hunched posture, rising liver enzymes, and drops in blood counts such as neutropenia, anaemia, or thrombocytopenia are all signs that an escalation break is needed. For immune-stimulating drugs, developers must also watch for fever, allergy-type reactions, low blood pressure, fast heart rate, and spikes in cytokines. Regular blood tests and body-weight tracking provide the practical dashboard to detect early organ stress, trigger temporary holds, and guide supportive care. The goal is not to eliminate all side effects – which is unrealistic for many cancer drugs – but to keep them predictable, manageable, and reversible within the study rules.
Top recommendations for sponsors (preventing pitfalls)
Four practices consistently separate smooth IND paths from those that stall.
- Anchor toxicology to the clinical plan. Set the target dose, expected exposure, and dosing schedule before you lock the GLP design so that the enabling study truly supports the trial you plan to run.
- Speed the process up with early characterisation. Short, hypothesis-driven tolerability and PK studies conducted before GLP will de-risk pivotal work by confirming schedule feasibility, exposure levels, and the need for rescue medications or additional monitoring.
- Protect your margins with validated methods. Begin pivotal studies only after TK and ADA methods have been validated. Unvalidated bioanalytical methods can undermine exposure-response interpretation and jeopardise first-cycle IND acceptance.
- Preserve study integrity. Resist the temptation to bolt new exploratory endpoints midstream. If an additional endpoint is important enough to consider, characterise it in pilots first so that the GLP package remains focused, interpretable, and submission-ready.
A final word
Oncology toxicology succeeds when speed and rigour work in tandem. Carefully planning the clinical schedule, making the best use of preclinical data specific to drug type, and meticulously calibrating the dosing, escalation, and breaking protocols are study-defining choices and must balance the need for safety with the urgent need for treatment. These choices, critical for any clinical study, but particularly nuanced for oncology drugs, compress time-to-IND without compromising scientific integrity or the safeguards that matter most to patients.
About the author
Tina Rogers, MBA, PhD, DABT, is senior technical director of toxicology at the WuXi AppTec Laboratory Testing Division. Her past leadership positions include vice president of preclinical sciences at Altasciences, executive vice president and director of research at M.P.I. Research, and vice president of drug development at Southern Research Institute.
