Phase 1 / 2 / 3 / 4 Trials Explained for Peptide Therapeutics

The short version

Clinical trials for peptide therapeutics are organized into phases that differ in size, population, and scientific question. Phase 0 / 1 trials involve tens to low hundreds of subjects and focus on pharmacokinetics and safety. Phase 2 trials involve hundreds of patients and ask whether the drug works and at what dose. Phase 3 trials involve thousands of patients and are the pivotal evidence for regulatory approval. Phase 4 trials run after approval to answer longer-term and broader-population questions. Understanding which phase a cited trial belongs to is the first step in calibrating how much weight to give its findings.

Phase 0 and microdosing

Phase 0 is an optional, non-therapeutic exploratory step, sometimes called a microdosing study. Subjects receive a single sub-therapeutic dose (typically 1/100th of the predicted pharmacologically active dose, not exceeding 100 micrograms) to generate early human PK data without meaningful pharmacological effect or toxicological risk.^[4]

For peptide therapeutics, Phase 0 studies are occasionally used when the sponsor wants to characterize absorption, distribution, and proteolytic degradation in humans before committing to full IND-enabling toxicology packages. They are particularly useful for comparing multiple peptide candidates head-to-head in humans with minimal regulatory burden. However, most peptide metabolic programs proceed directly from preclinical work to a conventional Phase 1 SAD/MAD design.

Phase 1: first-in-human safety and PK

Phase 1 trials typically enroll 10-100 subjects. For peptide metabolic drugs the population is usually healthy volunteers, which produces clean pharmacokinetic data without the confounders of polypharmacy or active disease pathology. In oncology, patients with the target cancer are enrolled from the outset because the experimental drug may offer direct benefit and the population cannot ethically be excluded.

The key outputs of Phase 1 for a peptide therapeutic are:

• Maximum tolerated dose (MTD) or maximum administered dose if no dose-limiting toxicity is reached.
• PK parameters: peak plasma concentration (Cmax), time to peak (Tmax), area under the curve (AUC), half-life (t½), clearance (CL), and volume of distribution (Vd).
• Initial adverse-event profile and dose-dependent tolerability pattern, which for GLP-1 receptor agonists is typically gastrointestinal (nausea, vomiting, diarrhoea).
• Evidence of target engagement: for a glucose-lowering peptide, even a healthy volunteer SAD study may include post-dose glucose or insulin measurements.

Phase 2: dose-finding and proof of concept

Phase 2 trials typically enroll 100-500 patients with the target condition and run for 12-26 weeks. The fundamental question is: which dose works best relative to placebo, and is the effect size large enough to justify a Phase 3 program?

Phase 2 trials are often divided into Phase 2a (proof of concept, smaller, exploring whether any signal exists) and Phase 2b (dose-ranging, larger, defining the Phase 3 doses). The distinction is informal; regulatory agencies do not use these labels in guidance documents, and a sponsor may combine the two into a single adaptive Phase 2 design.^[2]

A successful Phase 2 result for a peptide metabolic drug typically shows:

• A dose-response relationship across the arms tested (the drug's effect increases with dose, or plateaus at a defined maximum).
• Statistically significant improvement in the primary surrogate endpoint (e.g., HbA1c reduction) at one or more doses versus placebo.
• An acceptable adverse-event profile at the doses intended for Phase 3.
• PK-pharmacodynamic (PK/PD) modelling that supports once-weekly dosing or whatever regimen the sponsor plans to carry forward.

Phase 3: pivotal efficacy and the basis for approval

Phase 3 trials typically enroll 1,000-10,000 patients and are the primary basis for marketing authorisation. They are designed as adequate and well-controlled studies: randomized, usually double-blind or double-dummy, with a prespecified primary endpoint, a prespecified statistical analysis plan (SAP), and registration on a public trial registry before enrollment begins.^[1]

Reading a Phase 3 result table

Phase 3 publications in journals such as the New England Journal of Medicine or The Lancet follow a standardized structure. Here is what each element means:

• Primary endpoint. The single prespecified outcome on which the trial is powered to detect a difference. Statistical significance is tested at a two-sided alpha of 0.05 (p less than 0.05 rejects the null hypothesis). If the primary endpoint fails, the trial is considered a failure for regulatory purposes, regardless of secondary endpoint results.
• Key secondary endpoints. A hierarchical list of secondary endpoints tested in fixed sequence to protect the overall Type I error rate. Only secondary endpoints that are tested after the primary endpoint succeeds carry formal statistical inference; the rest are descriptive.
• Least-squares mean difference (LSMD). For continuous outcomes (HbA1c, body weight) the treatment effect is usually reported as the least-squares mean difference between the active arm and the placebo arm, with a 95% confidence interval (CI). If the CI excludes zero, the difference is statistically significant.
• Hazard ratio (HR) and 95% CI. For time-to-event outcomes (MACE, CV death, all-cause mortality), the treatment effect is reported as a hazard ratio. An HR of 0.80 means the event rate in the active arm is 20% lower than in the comparator arm at any given time point in the trial. An HR CI that excludes 1.0 is statistically significant; an HR upper bound below 1.30 satisfies the FDA non-inferiority criterion for CV safety.^[3]
• P value. The probability of observing the reported result (or a more extreme one) if the null hypothesis were true. A p value below 0.05 is the conventional threshold for statistical significance in Phase 3; it does not indicate the size or clinical importance of the effect.
• Number needed to treat (NNT). Not always reported in primary publications but calculable from response-rate data. The NNT is the number of patients who must receive the active treatment for one additional patient to achieve the outcome, compared with control.

The SURPASS-2 trial of tirzepatide illustrates this structure. The primary endpoint was HbA1c reduction from baseline at 40 weeks. All three doses of tirzepatide (5, 10, and 15 mg) were superior to semaglutide 1 mg in HbA1c reduction, with least-squares mean differences ranging from -0.45% to -1.13%. The key secondary hierarchical list included percent body weight change, fasting glucose, and proportion of patients achieving HbA1c below 7%.^[6]

Adaptive trial designs

An adaptive design allows prespecified modifications to the trial while it is ongoing, based on accumulated data, without compromising the validity of statistical inference. Common adaptations include:^[2][8]

• Combined Phase 2/3 design. The trial begins as a dose-finding study and transitions into a confirmatory pivotal trial once a dose has been selected, with data from both stages contributing to the final analysis. This shortens development timelines but requires careful Type I error control.
• Sample size re-estimation. An interim analysis reviews the observed effect size and adjusts the final sample size upward (but not downward) if the initial assumptions about variability or effect size were incorrect.
• Population enrichment. The trial may prespecify that enrollment will be restricted to a biomarker-defined subgroup if interim data show that the benefit is concentrated in that group.
• Basket and umbrella designs. Basket trials test one drug across multiple disease indications simultaneously; umbrella trials test multiple drugs within one disease, stratified by molecular or biomarker subtype. These designs are more common in oncology than in metabolic disease but are being explored for peptide therapeutics with multiple potential indications.

The FDA's 2019 guidance on adaptive designs requires that all adaptations, decision criteria, and statistical adjustments be specified in the protocol and SAP before the trial begins. Post-hoc adaptations based on unblinded data invalidate the statistical framework.

Phase 4 and real-world evidence

Phase 4 encompasses all study activity that occurs after a drug receives its initial marketing authorisation. The distinction from Phase 3 is regulatory context, not methodology: a Phase 4 study can be a double-blind randomized trial (for example, a cardiovascular outcomes trial required as a post-market commitment) or an observational registry study.

For peptide metabolic drugs, Phase 4 has delivered some of the most clinically consequential findings in the field: the LEADER and SUSTAIN-6 cardiovascular outcomes trials showed that liraglutide and semaglutide, respectively, reduced MACE in T2DM patients at high cardiovascular risk. These results, obtained in Phase 4 post-market commitments, expanded the clinical case for GLP-1 receptor agonists beyond glucose control.^[5]

Sample size across phases: a practical summary

These ranges are illustrative. Individual trials vary substantially based on effect size assumptions, required statistical power, endpoint variability, and regulatory requirements for the specific indication.