Preclinical vs Clinical Research Strategy for Drug Development

Less than 12 per cent of compounds that make it to Phase I clinical trials actually end up getting regulatory approval. Deciding to move a drug from preclinical tests to clinical research is therefore one of the toughest calls in pharmaceutical development.

At this point, the question is not whether a compound is “ready,” but whether the underlying evidence is sufficiently predictive of human outcomes to justify clinical risk. In an environment where Phase II failure remains structurally high, the quality of preclinical decision-making has become a primary determinant of R&D productivity.

Image

Core Distinctions Between Preclinical vs Clinical Research

Preclinical research tests a compound's safety and biological activity in the laboratory and in animal models before any human testing, whereas clinical trials test the compound in human participants under regulatory oversight across four formally designated phases.

The distinction matters because the two stages have fundamentally different objectives, regulatory frameworks, cost structures, and risk profiles. Conflating them, or using preclinical results to make unqualified assessments about clinical outcomes, is a primary driver of pipeline attrition.

* Tufts Center for the Study of Drug Development report estimates.

In summary, preclinical research is designed to de-risk human exposure; whereas clinical trials are designed to prove efficacy and safety in real people.

Image

What Preclinical Research Must Establish Before an IND Application

The Investigational New Drug (IND) application is the formal gateway between preclinical and clinical research. But from an executive standpoint, it is better understood as a risk conversion point, where relatively low-cost uncertainty becomes high-cost exposure.

Before a sponsor can submit an IND application to the U.S. Food and Drug Administration (FDA) or equivalent regulatory authority, preclinical research must generate a specific data package that addresses three core questions:

• Is the compound safe at clinically relevant exposure levels? Single-dose and repeat-dose toxicology studies in at least two species (typically one rodent, one non-rodent) must define the maximum tolerated dose and characterise target organ toxicity.
• Does the compound behave pharmacokinetically in a way that supports the intended route and frequency of administration? Absorption, distribution, metabolism, excretion, and toxicity (ADMET) profilling must demonstrate that the compound can reach the target tissue at therapeutic concentrations without unacceptable systemic exposure.
• Is there sufficient mechanistic evidence of biological activity to justify human risk? In vitro and in vivo pharmacology studies must show that the compound engages the intended target and produces the expected biological response at relevant dose levels.

These activities are often referred to collectively as IND-enabling studies, representing a defined milestone in the drug development process.

IND-enabling programmes integrate toxicology, pharmacokinetics, pharmacodynamics, and manufacturing readiness into a single, decision-critical package designed to support first-in-human dosing.

For leading pharma organisations, this phase is no longer treated as a regulatory formality, but rather as a strategic filter. The objective is not only to satisfy submission requirements but also to determine whether a compound has a sufficiently robust and predictive evidence base to justify entry into clinical development.

The FDA has 30 days to review an IND submission and place the application on clinical hold if the preclinical data package is insufficient.

In practice, the quality of preclinical evidence is the primary determinant of a smooth IND review. Regulators assess completeness. They do not assess predictive validity. That burden sits squarely with the sponsor and is where most strategic errors occur.

Preclinical Models: Strengths and Limitations in Predicting Clinical Outcomes

The translation gap is the disparity between results in preclinical models and outcomes in human trials. It is the most consequential challenge in drug development. Understanding why preclinical models succeed and fail at predicting clinical results is essential for interpreting preclinical data packages.

In vitro models

In vitro cell-based assays provide high-throughput, mechanistically defined data at relatively low cost. They are the first line of evidence for target engagement and cytotoxicity.

Their primary limitation is physiological reductionism: For example, a two-dimensional cancer cell line lacks the tumour microenvironment, vascularisation, and immune interactions that determine clinical drug response in vivo. 

In vivo animal models

Rodent and non-rodent animal models provide integrated systems data on pharmacokinetics, pharmacodynamics, and toxicology.

Image

Their principal limitation is species specificity: Human CYP enzyme metabolism, immune responses, and receptor pharmacology differ sufficiently from rodents that animal data cannot be directly extrapolated to human dose predictions.

Approximately 50–60% of Phase II failures are driven by lack of efficacy, much of which reflects insufficient predictive preclinical pharmacology and target validation. 

Computational models

Physiologically based pharmacokinetic (PBPK) modelling and in silico toxicity prediction tools have improved significantly and now contribute formally to IND data packages.

Their role is to fill gaps in experimental data, particularly for first-in-class mechanisms where animal models have limited predictive validity.

The FDA and European Medicines Agency (EMA) both accept PBPK modelling as supporting evidence in regulatory submissions, though not as a standalone substitute for experimental in vivo studies.

Clinical Trial Phases: What Each Stage Determines

Once an IND application is accepted, the clinical development programme proceeds through four sequentially gated phases, each with a distinct primary objective:

• Phase I: Assesses safety, tolerability, and pharmacokinetics in a small cohort (typically 20–100 participants). The primary endpoint is maximum tolerated dose. The success rate is approximately 60–65% across all therapeutic areas.
• Phase II: Assesses preliminary efficacy and dose optimisation in patients with the target indication (typically 100–500 participants). Phase II is the highest attrition point in clinical development, with an overall success rate of approximately 30–35%. Efficacy failure is the primary cause.
• Phase III: Aims to demonstrate confirmatory efficacy and safety in large, randomised controlled trials (typically 1,000–10,000 participants). Regulatory approval decisions are based primarily on Phase III data. The success rate is approximately 55–60% overall, with significant variation by therapeutic area.
• Phase IV: This encompasses post-marketing surveillance and pharmacovigilance. Phase IV continues indefinitely after approval and may result in label modifications, new indications, or, in rare cases, market withdrawal.

Image

The transition to first-in-human (FIH) studies represents the single highest uncertainty point in the drug development lifecycle. 

At this point, compounds move from controlled, model-based systems into human testing, where variability in genetics, disease state, and comorbidities can fundamentally alter drug response.

Critically, the risk at FIH is not limited to safety. It is the first real test of whether preclinical pharmacology translates into clinically meaningful exposure and biological effect.

As a result, the quality of IND-enabling evidence directly determines not only the safety profile observed in Phase I but also the probability that early clinical signals will be interpretable and actionable in later phases.

Where Preclinical Data Fails to Predict Clinical Outcomes

The translation gap is the technical term for the systematic divergence between preclinical findings and clinical results. It is not a failure of preclinical science per se but a consequence of the biological distance between model systems and human patients.

Three specific sources of translation failure account for the majority of Phase II attrition:

• Non-predictive animal models in CNS and oncology: Rodent models of Alzheimer's disease and certain cancers do not replicate the human disease mechanism with sufficient fidelity to predict clinical efficacy. Compounds that produce statistically significant effects in mouse models of Alzheimer's disease have a clinical success rate well below five per cent.
• Inadequate patient stratification at study entry: Preclinical studies test compounds in genetically homogeneous animal cohorts, whereas clinical trials enrol heterogeneous patient populations. Without validated predictive biomarkers to identify the patient subpopulation most likely to respond, Phase II trials often dilute a real treatment effect, leading to statistical insignificance.
• Dose translation errors: Converting the effective preclinical dose to a human starting dose requires validated allometric scaling or PBPK modelling. Errors in this calculation result in Phase I trials that test doses below or above the effective range, producing misleading safety and efficacy data.

The Real Strategic Decision: When Is Preclinical Evidence “Enough”?

For senior R&D leaders, the critical question is not whether the data meets regulatory requirements but whether the data supports confidence in a clinical hypothesis.

The FDA's IND guidance specifies the minimum data requirements for IND submission but does not prescribe a pass/fail threshold for preclinical evidence quality.

The sponsor's judgement about the strength of the preclinical case, particularly the mechanistic plausibility of the pharmacology and the relevance of the animal model, is therefore a critical strategic input that no regulatory checklist can fully replace.

Programmes that advance to clinical testing with weak mechanistic evidence, over-reliance on a single animal model, or unresolved ADMET liabilities carry a substantially higher probability of Phase II failure. The cost is not just financial: A failed Phase II trial delays the development of genuinely effective treatments for the patients who need them.

This requires moving beyond checklists toward multi-dimensional judgement:

• Is the mechanism of action validated across independent models?
• Are ADMET risks understood and proactively mitigated?
• Does the programme rely on a single line of evidence or a converging dataset?
• Is there a clear strategy for patient selection in early trials?

Advancing a compound without resolving these questions does not accelerate development. It shifts failure downstream, where it becomes exponentially more expensive.

From Stage Gates to Systems Thinking

The traditional view of preclinical-to-clinical transition as a linear progression is increasingly outdated.

Modern high-performing organisations are shifting toward:

• Integrated data environments linking preclinical and clinical insights
• Iterative learning models (e.g., DMTA cycles) that refine hypotheses continuously
• Early biomarker integration to align preclinical signals with clinical endpoints
• Predictive modelling frameworks that prioritise translational relevance over throughput

The objective is not to eliminate failure but to ensure that failure occurs early, cheaply, and informatively.

Conclusion: Preclinical Strategy Is Clinical Strategy

The boundary between preclinical research and clinical trials is no longer a simple handoff. It is a strategic checkpoint that determines whether a programme is built on predictive science or optimistic extrapolation.

In a landscape defined by rising development costs, increasing biological complexity, and persistent attrition, the organisations that outperform will not be those that move fastest into the clinic, but those that enter the clinic with the highest quality of evidence.

At Pharmatica, we focus on the decisions where scientific insight, operational execution, and strategic judgement intersect. Our analysis is designed to help pharmaceutical leaders identify where pipeline risk is truly created and how to reduce it before it compounds downstream.

Pharmatica: Insight. Connection. Impact.