Animal toxicology studies routinely surface findings that have no obvious human counterpart, from lesions in organs humans lack to rodent-specific tumor mechanisms and histopathology with no clinical marker. How FDA reads these findings affects IND clearance, the scope of clinical safety monitoring, the risk of a clinical hold, and what ends up in the label. Nonclinical, clinical and regulatory teams need to predict that response before they file.
The analysis below sets out the regulatory framework FDA applies, including the clinical hold criteria in 21 CFR 312.42 and the relevant nonclinical guidance. It then looks at the factors reviewers weigh: mechanistic relevance to humans, exposure margins, severity, reversibility, whether the effect can be monitored, and the patient population. It closes with precedents from FDA reviews where such findings were dismissed, monitored, used to restrict or hold a trial, or carried into labeling.
How FDA weighs nonclinical toxicity findings with no clear human correlate: clinical monitoring, clinical holds and labeling
Animal toxicology studies regularly produce findings that have no obvious human counterpart: lesions in organs humans do not have, tumors driven by rodent-specific mechanisms, histopathology with no clinical pathology marker, or effects whose mechanism is simply unknown. FDA does not treat "no human correlate" as a single category. Across guidance and review precedent, the agency's response turns on a small set of questions: Is the mechanism plausibly operative in humans? What is the exposure margin? How severe and how reversible is the finding? Can it be detected in patients before it becomes serious? And who is the patient population? The answers determine whether a finding is dismissed, monitored, used to restrict or halt a trial, or carried into labeling.
The regulatory framework
Clinical hold criteria
The legal threshold for a hold is set in 21 CFR 312.42. FDA may hold a Phase 1 investigation if subjects "are or would be exposed to an unreasonable and significant risk of illness or injury," if the investigator brochure is "misleading, erroneous, or materially incomplete," or if the IND "does not contain sufficient information required under § 312.23 to assess the risks to subjects" 560. For Phase 2 or 3 studies, those grounds apply, plus a clearly deficient protocol design 562. A hold "may apply to one or more of the investigations covered by an IND" 563, which is the basis for partial holds that restrict rather than stop development.
Two of these grounds matter most for unexplained animal findings. A finding can create an unreasonable and significant risk, or, where its cause has not been characterized, it can leave the IND without sufficient information to assess risk. The second ground is the one an unexplained animal death or an uncharacterized lesion most often triggers. FDA's oncology dosage-optimization guidance similarly notes that inadequate consideration of relevant nonclinical and clinical data may lead FDA to hold a protocol on risk, insufficient-information or deficient-design grounds 208.
ICH M3(R2): reversibility and monitorability as the core variables
ICH M3(R2) frames the purpose of nonclinical safety evaluation as characterizing toxic effects by "target organs, dose dependence, relationship to exposure, and, when appropriate, potential reversibility," and using that information "to identify parameters for clinical monitoring for potential adverse effects" 100. Two attributes recur throughout the guideline and its Q&As:
- Monitorability. A terminal recovery period is generally not warranted when toxicity can be monitored in humans early, is known to be irrelevant to humans, occurs only at exposures not considered clinically relevant, or is a manageable risk based on experience with related agents 104. A recovery study is generally not warranted when toxicity can be readily monitored in humans early, before it becomes severe 777879.
- Reversibility. Severe nonclinical toxicity with potential clinical impact calls for an evaluation of reversibility, which may be a scientific assessment rather than a dedicated recovery study. Anticipated nonreversibility belongs in the clinical risk assessment 777879.
These attributes also govern dose. For exploratory trials, M3(R2) permits a maximum dose up to one-half the NOAEL exposure in the more sensitive species "in cases where any relevant toxicity observed in animals is anticipated to be monitorable and reversible in humans" 101, and escalation above a specified AUC can be appropriate when findings are anticipated to be "monitorable, reversible, and of low severity in humans" 102. For standard Phase 1 trials, the maximum dose rests on risk assessments that include reversibility, monitorability, indication severity and observed clinical adverse effects 105. M3(R2) also recognizes that additional nonclinical studies "to identify potential biomarkers, to provide mechanistic understanding" can be useful when findings indicate special safety concerns 103.
Starting dose: nonmonitorable toxicity raises the safety factor
FDA's guidance on estimating the maximum safe starting dose makes the monitorability point concrete. Severe toxicities, nonmonitorable toxicities and significant toxicities that arise without reliable premonitory signs are reasons to apply a safety factor greater than the default 10. "Nonmonitorable toxicity" includes animal histopathologic changes that cannot readily be followed with clinical pathology markers 28. Irreversible toxicity is likewise listed because it suggests possible permanent injury to participants 2930. Conversely, a factor below 10 might be justified when toxicities are easily monitored, reversible and predictable 2930. In practice, a histopathology finding with no human biomarker is not ignored; it is managed by more conservative dosing.
Pediatric and oncology contexts
FDA's Nonclinical Safety Evaluation of Pediatric Drug Products guidance states that biomarkers of adverse effects found nonclinically may be used to monitor trial subjects, and that if a biomarker cannot be identified or used safely, an exposure level associated with toxicity may guide blood-level monitoring. If a juvenile-animal toxicity is likely in pediatric patients, cannot be monitored clinically and would be an unacceptable consequence of exposure, "it may not be possible to conduct pediatric trials safely" 81. That is the clearest guidance statement linking nonmonitorability to a trial not proceeding.
For advanced cancer, ICH S9 relaxes these expectations: a NOAEL is not essential, Phase 1 can be driven by MTD and dose-limiting toxicity, and the highest nonclinical exposure generally does not cap clinical escalation 112115. An assessment of whether serious adverse effects are reversible is still expected 113114.
Species relevance: carcinogenicity and biotechnology products
ICH S1's weight-of-evidence approach treats a rodent tumor as something to investigate for human relevance, not as an automatic prediction of human cancer. The carcinogenicity assessment considers target biology in rats and humans, genotoxicity, chronic histopathology and mechanism, metabolic differences, hormonal effects and exposure 233234. Category 3a covers compounds likely to cause rat tumors through "prior established, well-recognized mechanisms known to be human irrelevant" 235. The guideline gives hepatic enzyme induction with rat hepatocellular hypertrophy as an example of a rodent-specific tumor mechanism, but expects the sponsor to show the data supporting the mechanism rather than merely assert irrelevance 236. Where mechanism, exposure comparison or human relevance is inadequately supported, the uncertainty itself can make a rodent study valuable 235239.
For biologics, ICH S6(R1) makes pharmacological relevance of the test species the starting point, warns that studies in non-relevant species may be misleading 434445, and treats animal anti-drug antibody findings as tools to interpret toxicity data rather than predictors of human immunogenicity 4849.
How the variables combine
The framework above, and the precedents that follow, reduce to a reasonably consistent pattern.
| Finding profile | Typical FDA response |
|---|---|
| Mechanism shown to be species-specific, or lesion in an organ humans lack | Not dose-limiting; often no specific action |
| Human relevance unknown, adequate exposure margin, not severe | Described in labeling (Section 13.1 or 13.2), often with "clinical significance unknown" |
| Human relevance unknown, potentially serious, matters to treatment decisions | Warnings and Precautions, occasionally a boxed warning |
| Potentially serious and monitorable with an available marker | Targeted monitoring in trials; may carry into labeled monitoring instructions |
| Serious, unexplained (for example, animal deaths of unknown cause) or not monitorable | Clinical hold until characterized, or partial hold restricting population or dose |
Precedents: clinical holds
| Drug | Nonclinical finding | Hold and resolution |
|---|---|---|
| Golodirsen (Vyondys 53) | Unexpected deaths in a toxicology study | Full clinical hold; the sponsor was asked to investigate the cause. The hold was lifted after data identified renal toxicity as the cause of death in high-dose animals, which the reviewer considered monitorable in humans 140 |
| Tadalafil (Cialis) | Lung perivascular inflammation, coronary lesions and disseminated arteritis in 1- and 6-month dog studies 141481482483 | Hold placed on the IND in December 1997 484; FDA requested a 12-month dog study 143 |
| Mipomersen (Kynamro) | Vasculitis and intimal hyperplasia after a year of dosing in monkeys | Partial clinical hold limiting study to high-risk patients 144145; the pre-NDA minutes record that the partial hold remained in place 146 |
| Ensitrelvir (Xocova) | Deaths in monkeys in a two-week repeat-dose study, with low safety margins | Hold lifted in March 2022 after a lower-dose protocol and additional clinical safety data 147148 |
Two lessons stand out. First, an unexplained finding is more likely to trigger a hold than an understood one. In golodirsen, the hold ended not because the toxicity disappeared but because its cause was identified and judged monitorable 140. Second, holds can be calibrated. The mipomersen partial hold limited exposure to a population with a more favorable benefit-risk balance rather than stopping the program 144145, and the ensitrelvir hold was resolved partly through dose reduction 147148.
The tadalafil file shows how a species-specificity argument is tested. The sponsor invoked beagle pain syndrome, a spontaneous canine vasculitis. The reviewer accepted that mild coronary lesions, including lesions in controls, could reflect background incidence, but remained concerned about moderate-to-severe disseminated arteritis in treated dogs 485. Vascular findings also appeared, at lower severity, in mice and rats, and no specific biomarker was identified 486487. The 12-month dog study did not reproduce disseminated arteritis 143, and the sponsor was asked to examine ESR and CK in patients; subsequent studies found no clinically significant elevations 488489. The animal findings were nonetheless disclosed in the Animal Toxicology section of labeling, including the contrasting 12-month results 490.
Precedents: clinical monitoring derived from animal findings
When an animal finding is plausible in humans and detectable, FDA's preference is to monitor rather than halt.
| Drug | Nonclinical finding | Monitoring | Human outcome |
|---|---|---|---|
| Paltusotine (Palsonify) | Unilateral corneal dystrophy and focal retinopathy in rats, not dose-related and unlike typical phototoxicity 585 | FDA requested visual acuity, macular photography, OCT and visual fields at baseline and every six months in trials of at least six months 586 | Ocular adverse events similar to placebo (8.3% vs 8.2%) in placebo-controlled phases 587; later retinal findings could not be reliably attributed because monitoring began after database lock without baseline exams 590591 |
| Safinamide (Xadago) | Retinal degeneration in rodents; retinal scarring and cataracts in rats | Close ophthalmologic monitoring including OCT and electroretinography 173 | No eye-damage signal, but limited and missing examinations meant the data could neither rule out nor confirm the animal risk 173 |
| Fingolimod (Gilenya) | Lung toxicity in animals 174 | Pulmonary function testing including FEV1 and DLCO 175 | Dose-related decreases, including at the recommended dose 176 |
| Mavorixafor (Xolremdi) | Liver histopathology and elevated liver tests in animals | Liver function monitoring incorporated into the WHIM syndrome study 178 | Not reported in the FDA review |
| Edoxaban (Savaysa) | Prolonged ocular retention in animals | Close ocular monitoring of repeatedly dosed subjects recommended 180 | Not reported in the FDA review |
| Golodirsen (Vyondys 53) | Dose-related, predominantly tubular kidney injury; renal failure and deaths in juvenile rats; possible irreversible damage with a small exposure margin 396397398 | Renal biomarkers including BUN, cystatin C and KIM-1 assessed in trials 400377 | Mild, transient proteinuria cases; no serious renal toxicity 405406407 |
The paltusotine and safinamide reviews illustrate a recurring limitation: monitoring that starts late or is incomplete may leave the human relevance question unanswered at the time of the marketing decision. For paltusotine, the review stated that the nonclinical and clinical ocular findings would be included in labeling and relied on enhanced pharmacovigilance for ocular phototoxicity after marketing 592. Ophthalmology reviewers noted that continued trial monitoring and/or a postmarketing study could give a more definitive answer 593594.
Golodirsen shows monitorability itself being contested. A 2019 complete response letter raised concern that low creatinine production in patients with DMD made kidney injury hard to monitor. On appeal, the OND Director concluded that serial renal measures could be used and called for baseline direct GFR and specific labeled monitoring instructions 399. The final label's Warning 5.2 calls for baseline cystatin C, urine dipstick and urine protein-to-creatinine ratio, consideration of measured GFR, monthly dipstick, quarterly cystatin C and UPCR, and nephrology referral for persistent abnormalities, and cautions that creatinine may be unreliable in DMD 402403. The same label states that kidney toxicity was not observed in the limited clinical studies 404. This is a clear example of an animal-only finding producing labeled monitoring requirements.
Precedents: labeling findings of uncertain human relevance
Regulatory basis
Under 21 CFR 201.57, Section 13.1 must report the conduct and results of carcinogenicity studies and animal findings raising mutagenesis or fertility concerns, and Section 13.2 must include "significant animal data needed for safe and effective human use" not incorporated elsewhere 414415. Warnings and Precautions covers clinically significant hazards, and a warning is required once there is reasonable evidence of a causal association 413. FDA's labeling guidance on Warnings and Precautions adds that an animal finding should be included there when it raises substantial concern that a serious or otherwise clinically significant adverse reaction could occur in humans, even if not observed with the drug, and that labeling should say the reaction has not been observed but is anticipated when that is important to prescribers 298299. The pediatric guidance likewise allows labels to include nonclinical adverse effects not demonstrated clinically, and to restrict or warn against pediatric use depending on the findings 436369.
Section 13: disclosure without a warning
The most common outcome for an unexplained finding at an adequate margin is disclosure in Section 13 with an explicit statement of uncertain significance.
| Drug | Finding | Placement and reasoning |
|---|---|---|
| Adagrasib (Krazati) | Phospholipidosis in animals | Section 13.2; human significance unknown 325326 |
| Relugolix (Orgovyx) | Phospholipidosis in rats and monkeys | Section 13.2; human significance unknown and not observed in the clinical studies cited 327 |
| Capmatinib (Tabrecta) | Rat brain white-matter vacuolation, in some animals with convulsions and early death; partially reversible 505 | Section 13.2; FDA considered it relevant enough to disclose because of its frequency and potential effects near clinical exposure, although whether it occurs in patients is unknown 329. The label notes onset at ≥2.2 times the human AUC 506 |
| Sonidegib (Odomzo) | Tremors with occasional CK elevations; dog tremors above the MTD | Section 13.2; pharmacological relevance uncertain 330 |
| Silodosin (Rapaflo) | Reversible reduced fertility in male rats, also seen with class members | Included in labeling with a statement that clinical relevance is unknown; species specificity was not established 268 |
| Oteseconazole (Vivjoa) | Cataracts in rats exposed during pre- and postnatal development | Recommended for the Animal Data portion of labeling although reviewers could not determine human risk 334 |
Carcinogenicity findings follow the same logic. For bempedoic acid (Nexletol), FDA proposed describing male rat pancreatic tumors with a statement that clinical relevance was unclear, despite the applicant's argument that they were human-irrelevant 339. For buprenorphine extended-release (Brixadi), FDA said mouse tumors associated with the NMP excipient must be included because the mode-of-action assessment did not adequately establish human irrelevance, even though an eightfold margin suggested the human risk might not be significant 340. In these files the burden sits with the sponsor: an asserted mechanism that is not adequately supported does not remove a finding from labeling.
Escalation to Warnings and Precautions or a boxed warning
Some unexplained findings are judged important enough to reach prescribers outside Section 13.
Flibanserin (Addyi). A two-year study showed a dose-related increase in malignant mammary tumors in female mice at about 3 and 10 times clinical exposure, not seen in male mice or rats 459460. A prolactin mechanism was considered unlikely and genotoxicity did not explain it, so the mechanism and human relevance remained unknown 462. One reviewer withdrew an earlier recommendation to contraindicate use in women with breast cancer or a family history and proposed Section 13.1 only 463. The cross-discipline team leader cautioned that an unknown mouse mechanism meant a negative rat study could not exclude clinical risk in healthy premenopausal women who might use the drug chronically 464. The nonclinical supervisor recommended Warnings and Precautions as well as 13.1 because the finding could affect treatment decisions and might be missed in the animal-toxicology section 332. The current label carries Warning 5.7, "Mammary Tumors in Female Mice," stating that clinical significance is unknown 465. The population (healthy individuals, chronic use, an elective indication) drove the escalation as much as the tumor data.
Exenatide extended-release (Bydureon). Rat thyroid C-cell tumors occurred at clinically relevant exposures, and labeling states that the relevance of the rodent findings has not been determined 552. FDA nonetheless required a boxed warning 553, a contraindication in patients with a personal or family history of MTC or MEN 2 554, and a REMS addressing the potential MTC risk along with acute pancreatitis 556. The label also illustrates the limits of monitoring: routine calcitonin measurement or thyroid ultrasound is described as of uncertain value and potentially leading to unnecessary procedures 552555. Where a credible surrogate does not exist, FDA relied on restriction and communication rather than monitoring.
Precedents: findings accepted as species-specific
FDA does accept species-specificity arguments when the anatomy or mechanism supports them.
- Revefenacin (Yupelri). Laryngeal epithelial inflammation was called rodent-specific and not relevant to humans, and Harderian gland pigmentation was deemed clinically irrelevant because humans have no analogous structure. Neither was treated as dose-limiting 262.
- Bempedoic acid (Nexletol), juvenile toxicity. Increased liver weight and hepatocellular hypertrophy in juvenile rats were attributed to rodent-specific PPARα effects, so the absence of an exposure margin was not considered a clinical concern 263. The pancreatic tumor finding in the same program, by contrast, went into labeling 339.
- Cobicistat (Tybost). Rat thyroid follicular tumors were described as rat-specific, driven by PXR activation, hepatic CYP3A induction and thyroid hormone imbalance, a mechanism the reviewer said does not occur in humans 264.
- Palovarotene (Sohonos). Forestomach degeneration and inflammation were considered not clinically relevant because humans lack a comparable structure 265266. Separately, a postmarketing requirement for rodent carcinogenicity studies was recommended because human cancer risk remained uncertain 254267, showing that dismissing one finding does not settle the broader safety question.
A cautionary reversal: roflumilast
Roflumilast (Daliresp) caused nasal epithelial tumors in hamsters but not mice 528529. FDA's Executive Carcinogenicity Assessment Committee initially considered the tumors potentially not relevant to humans because they were attributed to the metabolite ADCP N-oxide, which hamsters but not humans were believed to produce; even then, the committee said the finding should be noted in labeling 530. FDA later reversed the rodent-specific determination after new human pharmacokinetic data detected ADCP N-oxide in plasma and urine, leaving human relevance unknown 531. A species-specificity argument that depends on a metabolic difference is only as durable as the human metabolite data behind it.
Practical implications for sponsors
- Characterize unexplained findings before FDA has to. Holds in these precedents followed unexplained deaths or serious lesions. Identifying the cause, as in golodirsen, was the route out 140. M3(R2) explicitly supports mechanistic and biomarker studies for this purpose 103.
- Build the monitorability case explicitly. Identify the clinical marker, its sensitivity relative to the animal lesion, and the point at which it would detect injury before it becomes severe. Where no marker exists, expect a higher starting-dose safety factor 28 or, in pediatric programs, questions about whether trials can proceed safely 81.
- Start monitoring at baseline. Paltusotine and safinamide show that late or incomplete monitoring leaves the human relevance question open at approval 590591173.
- Support species-specificity claims with data, not assertion. ICH S1 expects the mechanism to be demonstrated 236. FDA labeled bempedoic acid pancreatic tumors and the Brixadi excipient tumors when mechanism data were judged inadequate 339340, and reversed roflumilast when human metabolite data emerged 531.
- Expect population and indication to shape placement. Flibanserin's warning reflected use in healthy women 332464; S9 tolerates more for advanced cancer 112115; mipomersen's partial hold restricted study to high-risk patients 144145.
- Plan for "clinical significance unknown" language. For findings at adequate margins with uncertain mechanism, Section 13 disclosure with that qualifier is the default outcome across a wide range of recent reviews 325327329268.
Deeper follow-ups, such as how FDA has handled specific organ toxicities (phospholipidosis, retinal findings in albino rodents, crystalluria) or how these decisions compare with EMA practice, can be explored further by asking Rhizome directly.