Clinical Trial Terminology

ARR = control event rate - treatment event rate. ARR directly tells you what proportion of patients benefit from treatment and is the basis for NNT calculation. ARR is population-dependent — the same RRR produces different ARRs in high-risk vs low-risk populations. For GLP-1 RA trials, the ARR for weight achievement thresholds (≥5%, ≥10%, ≥15% weight loss) can be large: if 85% of treated patients vs 30% of placebo patients achieve ≥5% weight loss, the ARR is 55% (NNT ≈ 2). This contrasts with the smaller ARR for preventing cardiovascular events (~1-2% over 3-5 years). ARR contextualises clinical benefit more directly than RRR for individual patient decision-making.

Active comparator designs answer a more clinically relevant question than placebo trials: is the new drug better than, equal to, or not worse than the current standard? Semaglutide Phase III trials used active comparators including sitagliptin (SUSTAIN 2), exenatide ER (SUSTAIN 3), insulin glargine (SUSTAIN 4), and dulaglutide (SUSTAIN 7). Active comparator trials require larger sample sizes than placebo-controlled trials because the comparator also produces a treatment effect, narrowing the between-group difference. The choice of comparator affects interpretation — comparing against a weak comparator inflates apparent benefit, which is why regulators and guidelines committees scrutinise comparator selection.

Adaptive designs pre-specify interim analyses that can modify: sample size (based on observed treatment effect), randomisation ratio (shift more patients to promising arms), dose levels (drop futile arms, add new arms), patient population (enrich for responders), endpoints, or trial duration. These modifications must be planned before the trial starts and defined in the statistical analysis plan — post hoc changes are not adaptive design. Bayesian adaptive designs use accumulating data to update probability distributions. Adaptive platform trials (e.g. I-SPY for cancer) can continuously add and drop treatments. For peptide drug development, adaptive Phase II designs can efficiently identify optimal doses from multiple candidates while controlling overall Type I error rate.

AE reporting in clinical trials follows MedDRA (Medical Dictionary for Regulatory Activities) terminology — a standardised coding system with System Organ Classes (e.g. gastrointestinal disorders), preferred terms (e.g. nausea), and lower-level terms (more specific descriptions). AEs are graded by severity (mild, moderate, severe) and assessed for causality relationship (related, possibly related, unlikely related, not related). Key summary metrics: incidence (proportion experiencing at least one event), exposure-adjusted incidence rate (events per patient-year of exposure), and treatment-emergent AEs (new or worsened after first dose). For GLP-1 RAs, gastrointestinal AEs (nausea 15-44%, vomiting 5-25%, diarrhoea 8-20%, constipation 5-15%) are the most common and drive most discontinuations.

Basket trials test a single treatment across multiple disease subtypes that share a common molecular feature. For example, a somatostatin analogue could theoretically be tested in a basket trial across different NET subtypes (pancreatic, small bowel, pulmonary) that all overexpress SSTR2. The statistical challenge is that efficacy may vary across subtypes — basket trials must balance testing within each basket against borrowing strength across baskets. Bayesian hierarchical models can share information between baskets while allowing for heterogeneity. Basket designs are most common in oncology but the concept applies to any drug targeting a specific receptor or pathway expressed across multiple conditions.

Blinding challenges for peptide drug trials: injectable drugs require matched placebo devices (same pen type, appearance, injection procedure), different injection frequencies between drugs require dummy dosing schedules (double-dummy design — e.g. comparing daily vs weekly injection requires the daily group to receive weekly placebo and the weekly group to receive daily placebo), and distinctive side effects (GLP-1 RA nausea) may functionally unblind participants. Blinding assessment: some trials include questionnaires asking participants to guess their treatment assignment — if guessing accuracy significantly exceeds 50%, functional unblinding may have occurred. Central adjudication of endpoints (blinded committee reviews de-identified data) provides an additional layer of bias protection regardless of participant-level blinding.

The FDA mandated CVOTs for new diabetes drugs following the rosiglitazone cardiovascular safety concern (2008 guidance). CVOTs must demonstrate that the drug does not increase cardiovascular risk (upper bound of 95% CI for HR <1.3 for pre-approval; <1.8 post-marketing). Notable peptide CVOTs: LEADER (liraglutide) — HR 0.87 for MACE, first GLP-1 RA to show CV benefit; SUSTAIN-6 (semaglutide 0.5/1.0mg) — HR 0.74 for MACE; SELECT (semaglutide 2.4mg in obesity without diabetes) — HR 0.80 for MACE; SURPASS-CVOT (tirzepatide) — ongoing. These CVOTs transformed GLP-1 RA positioning from glucose-lowering agents to cardioprotective drugs, with treatment guidelines now recommending them for patients with established cardiovascular disease.

Case reports follow the CARE (CAse REport) guidelines for reporting quality and are typically structured as: introduction, patient information, clinical findings, diagnostic assessment, therapeutic intervention, follow-up/outcomes, and discussion. In the peptide research space, case reports may document: unexpected adverse effects of approved peptide drugs, novel off-label applications, unusual drug interactions, or clinical observations in patients using research compounds. While case reports cannot establish causality, they have historically been the source of many important drug safety signals. The BMJ Case Reports and Journal of Medical Case Reports are prominent outlets. PubMed indexes case reports alongside other publication types.

Case series document similar clinical observations in multiple patients (typically 3-50+) and can provide preliminary evidence of a treatment effect or safety signal that justifies formal investigation. They are retrospective (reviewing medical records) or prospective (collecting data as patients present). Case series lack a control group, making it impossible to attribute outcomes to the treatment rather than natural disease course, placebo effect, or other factors. For research peptides without clinical trial data, case series published in peer-reviewed journals represent the strongest available human-level evidence (though still far below RCT evidence). Quality assessment tools (IHE case series checklist) help evaluate the reliability of case series data.

Case-control studies start with the outcome (cases) and look backward for exposures. They are particularly efficient for studying rare events — instead of following thousands of patients to observe rare events prospectively, researchers identify individuals who already experienced the event and compare their treatment history to matched controls. For peptide drug safety, case-control designs have been used to investigate: thyroid cancer in GLP-1 RA users (cases with thyroid cancer matched to controls without), pancreatitis associated with incretin-based therapies, and rare adverse events with specific antimicrobial peptides. Odds ratios (the case-control equivalent of relative risk) quantify the association. Recall bias (cases may recall exposures differently) and selection bias (control selection) are key limitations.

Clinical endpoints directly measure outcomes that matter to patients: survival (overall survival — the gold standard), disease events (heart attack, stroke, fracture), functional measures (6-minute walk distance, exercise capacity), and symptom resolution (pain scores, urinary frequency). Clinical endpoints are preferred over surrogate endpoints for regulatory decision-making but may require very large trials (thousands of patients) followed for years to accrue sufficient events. The cardiovascular outcomes trials for GLP-1 RAs used clinical endpoints (MACE events) but required 3,000-17,000 patients followed for 2-5 years to demonstrate statistically significant differences in cardiovascular event rates.

Clinical trials follow a regulatory framework established by ICH-GCP (International Council for Harmonisation Good Clinical Practice) guidelines. Key requirements include: written protocol specifying objectives, design, methodology, and statistical analysis plan; ethics committee/IRB approval before enrolment; informed consent from all participants; data integrity through case report forms and electronic data capture; monitoring by the sponsor or contract research organisation (CRO); adverse event reporting to ethics committees and regulatory authorities; and trial registration on public databases (ClinicalTrials.gov, EU Clinical Trials Register). The drug development process from preclinical through Phase III typically takes 10-15 years and costs $1-3 billion. Only approximately 12% of drugs entering Phase I eventually receive approval.

The database contains over 475,000 study records from 220+ countries. The API (v2) allows programmatic access to trial data using compound name queries, condition filters, status filters, and other parameters. Results reporting requirements (FDAAA 801): sponsors must submit summary results within 12 months of primary completion date for applicable clinical trials. Results include participant flow, baseline characteristics, outcome measures, and adverse events. The EU Clinical Trials Register (EudraCT) contains European trial data and interfaces with ClinicalTrials.gov. WHO's International Clinical Trials Registry Platform (ICTRP) aggregates trial records from multiple national registries. Together, these registries enable comprehensive tracking of clinical trial activity for any compound.

Prospective cohorts follow participants forward from exposure to outcome (higher quality, can establish temporal sequence, but slow and expensive). Retrospective cohorts use historical data (faster and cheaper, but limited by available data quality). Large healthcare database cohort studies using electronic medical records (EMR), insurance claims, and national registries have become important for peptide drug post-marketing safety surveillance. Examples: cohort studies using Scandinavian national health registries examined long-term outcomes of GnRH agonist therapy; insurance claims databases tracked GLP-1 RA users for pancreatic and thyroid safety signals. Cohort studies cannot eliminate confounding (observational, not randomised) — propensity score matching and instrumental variable methods partially address this limitation.

MACE (major adverse cardiovascular events) is the standard composite endpoint for cardiovascular outcomes trials: 3-point MACE = cardiovascular death + non-fatal myocardial infarction + non-fatal stroke; 4-point MACE adds hospitalisation for unstable angina. Composite endpoints increase event rates (improving statistical power with smaller samples/shorter follow-up) but individual components may not contribute equally and may even move in different directions. The SELECT trial (semaglutide 2.4mg for cardiovascular risk reduction in obesity) showed a 20% reduction in 3-point MACE, with the benefit driven primarily by non-fatal MI and non-fatal stroke reduction. Careful analysis of individual components is essential when interpreting composite results.

A 95% CI means: if we repeated this experiment many times, approximately 95% of the calculated intervals would contain the true population parameter. The width of the CI reflects precision — narrower intervals indicate more precise estimates (larger sample sizes produce narrower CIs). For clinical decision-making, CI lower and upper bounds are more informative than p-values alone. If a 95% CI for weight loss difference excludes zero, the result is statistically significant at p<0.05. The clinical interpretation considers whether the entire CI represents clinically meaningful values. For non-inferiority trials, the CI must lie entirely above the pre-specified non-inferiority margin.

Crossover designs require: stable chronic condition (not progressive or self-limiting), drug effects that are fully reversible during the washout period (no carryover effect), and sufficient washout duration (typically 5-7 half-lives). Statistical analysis accounts for period effects (treatment may differ in first vs second period) and carryover effects (residual effect from first period). AB/BA designs (simplest crossover) randomly assign half to treatment A→washout→B and half to B→washout→A. For peptide drugs with very long half-lives (weekly semaglutide t1/2 ~1 week would require ~5-7 week washout), crossover designs may be impractical. Crossover designs are most commonly used in bioequivalence studies and PK comparisons.

DSMBs (also called Data Monitoring Committees, DMCs) typically include: clinical experts in the disease area, at least one biostatistician, and sometimes an ethicist. DSMB members must be independent of the sponsor and investigators. They review unblinded data (the only people permitted to see treatment assignments during the trial) at pre-specified interim analyses. DSMB actions include: continue as planned, modify the protocol, temporarily halt enrolment, or permanently stop the trial (for overwhelming benefit — ethical to make treatment available sooner; for safety — unacceptable harm; or for futility — treatment unlikely to show benefit even if continued). DSMB deliberations are confidential. The DSMB charter defines operating procedures, interim analysis timing, and statistical stopping boundaries.

The traditional 3+3 design: 3 patients at each dose level; if 0/3 DLTs → escalate; if 1/3 DLTs → add 3 more patients (if 1/6 DLTs → escalate, if ≥2/6 → MTD exceeded, previous dose = MTD); if ≥2/3 DLTs → MTD exceeded. Limitations: treats MTD as a fixed threshold, uses limited data at each level, tends to recommend doses below the true MTD. Model-based designs (continual reassessment method, CRM; Bayesian optimal interval, BOIN) use accumulating data to continuously update the dose-toxicity relationship and target a specific DLT rate (usually 20-30%). Accelerated titration designs (single patient per cohort until first toxicity) can speed initial escalation through clearly sub-therapeutic doses. Peptide-specific considerations: immunogenicity may be dose-dependent and delayed, requiring extended observation windows.

DLTs are pre-defined in the trial protocol and typically include: Grade 3-4 non-haematological toxicity, Grade 4 haematological toxicity lasting >7 days, febrile neutropenia, any toxicity resulting in dose delay >2 weeks, or any toxicity considered dose-limiting by the investigator. DLT assessment periods are defined (typically the first cycle or first 28 days). The MTD is often defined as the highest dose at which <33% of participants experience a DLT (the 3+3 dose escalation design) or using more sophisticated model-based methods (Bayesian continual reassessment method, BOIN design). For peptide drugs, common DLTs include gastrointestinal intolerance (GLP-1 RAs), injection site reactions, immunological reactions, and target-related pharmacological effects that become excessive.

Phase IIb dose-ranging studies randomise patients to 3-5 dose levels plus placebo. Design considerations: dose spacing (geometric progression, e.g. 1mg, 3mg, 10mg, 30mg — covering a wide range with equal ratios), sample size per arm (typically 30-100 patients, balancing precision against cost), and treatment duration (long enough to observe near-maximal effect). Analysis: dose-response modelling using Emax models (hyperbolic curve fitting), linear or log-linear models, or more flexible approaches (MCP-Mod — multiple comparisons procedure combined with modelling). The selected Phase III dose should be on the plateau of the dose-response curve (near-maximal efficacy) with acceptable tolerability. Tirzepatide Phase II identified 5mg, 10mg, and 15mg as the Phase III doses from a broader range tested.

Maintaining blinding for injectable peptide drugs requires careful matching: the active drug and placebo must be in identical devices (same pen type, cartridge appearance, injection volume, needle gauge), the solutions must have the same appearance (colour, viscosity), and injection-related sensations should be similar (injection site reactions could theoretically unblind, which is why ISR rates are carefully compared between groups). In-trial blinding can be compromised by distinct pharmacological effects — nausea with GLP-1 RAs may indicate active treatment to perceptive participants or investigators. To mitigate this, trials may use matching placebos with added components to produce similar mild effects (rarely done), or rely on centralised outcome assessment by blinded adjudication committees to remove investigator bias.

Effectiveness-efficacy gaps arise from: variable adherence (missed doses, early discontinuation — GLP-1 RA 12-month persistence rates range from 30-70% in real-world data vs >90% in trials), broader patient populations (trial exclusion criteria screen out complicating comorbidities), less intensive monitoring (trial visits every 4-12 weeks include motivational support that routine care doesn't match), and concomitant medication changes (real-world polypharmacy may differ from controlled trial protocols). Pragmatic trials (designed to mirror real-world practice) bridge the efficacy-effectiveness gap. Real-world effectiveness studies of GLP-1 RAs consistently show lower HbA1c and weight reduction than trials, primarily driven by adherence — patients who persist on therapy achieve outcomes close to trial efficacy.

Efficacy assessment requires controlled conditions to isolate the drug's effect from confounders, placebo effect, and regression to the mean. Key efficacy metrics vary by indication: diabetes (HbA1c change, fasting glucose, time in range), obesity (% weight loss, proportion achieving ≥5/10/15/20% loss), cancer (overall survival, progression-free survival, response rate), osteoporosis (fracture incidence, BMD change), and hormonal conditions (hormone levels, symptom scores). The treatment policy estimand (ITT) reflects efficacy as prescribed (including dropouts), while the trial product estimand (completers) reflects efficacy with adherence. The difference between these estimates quantifies the impact of non-adherence. Tirzepatide 15mg achieved up to 22.5% mean weight loss (ITT, treatment policy) and 26.6% (completers), demonstrating both high efficacy and the importance of estimand specification.

Equivalence designs use two-sided testing: the 95% CI of the treatment difference must fall entirely within the pre-specified equivalence margins (-delta to +delta). This is more stringent than non-inferiority (which only tests one direction). Equivalence trials are used for: biosimilar comparisons (demonstrating the biosimilar is neither better nor worse than the reference product), bioequivalence studies (comparing PK parameters of generic vs brand formulations), and head-to-head comparisons where the goal is to demonstrate comparable efficacy. For peptide biosimilars, clinical equivalence trials typically use the same primary endpoint as the reference product's pivotal trial, with margins determined by reference to historical treatment differences.

EudraCT (European Union Drug Regulating Authorities Clinical Trials Database) is maintained by the EMA and contains information on all clinical trials conducted in the European Economic Area since 2004. The number format is YYYY-NNNNNN-CC where YYYY is year, NNNNNN is a sequential number, and CC is a check digit. EudraCT records include: protocol information, regulatory and ethics approvals, and (since 2014) summary results for paediatric and Phase II-IV adult trials. The EU Clinical Trials Register (clinicaltrialsregister.eu) is the public-facing portal. Under the new Clinical Trials Regulation (CTR 536/2014), the Clinical Trials Information System (CTIS) is replacing EudraCT as the single submission point for EU clinical trials.

Common exclusion criteria for peptide drug trials include: pregnancy or planned pregnancy (most peptide drugs are contraindicated), breastfeeding, severe renal impairment (eGFR <30 mL/min — affects peptide clearance), severe hepatic impairment, known hypersensitivity to the drug or class, history of pancreatitis (for GLP-1 RAs), personal/family history of medullary thyroid carcinoma or MEN2 (for GLP-1 RAs), type 1 diabetes (for type 2 diabetes trials), bariatric surgery within specified timeframe, recent use of competing therapies (washout required), psychiatric illness affecting consent capacity, and participation in another clinical trial. Exclusion criteria protect participant safety but reduce generalisability.

ICH E6(R2) GCP covers: institutional review board responsibilities, investigator qualifications and obligations, sponsor responsibilities, clinical trial protocol requirements, investigator's brochure content, essential documents, and quality management systems. Key GCP principles: trials should be scientifically sound, protection of participant rights/safety/wellbeing takes priority over science/society, adequate preclinical evidence must support clinical use, protocols must be reviewed by ethics committees, informed consent is essential, every participant must be under qualified medical care, adverse events must be reported, data must be accurately recorded and verifiable against source documents, and participant confidentiality must be maintained. GCP inspection by regulatory authorities verifies compliance and data integrity.

The HR is derived from Cox proportional hazards regression and reflects the instantaneous risk ratio between groups. HR = 0.80 means the treatment group has 20% lower hazard (instantaneous risk) of the event at any given time point. Key assumptions: proportional hazards (the HR remains constant over time) and non-informative censoring (censored patients don't differ systematically from those remaining). The SELECT trial reported HR 0.80 (95% CI 0.72-0.90) for 3-point MACE with semaglutide 2.4mg, indicating a 20% cardiovascular risk reduction. Kaplan-Meier curves visualise the event-free survival over time and allow assessment of whether the proportional hazards assumption holds (curves should not cross).

Inclusion criteria define the target population and ensure enrolled participants can meaningfully contribute to the trial's objectives. For GLP-1 RA diabetes trials (example SUSTAIN programme): adults ≥18 years, type 2 diabetes diagnosis, HbA1c 7.0-10.5%, on stable background metformin for ≥30 days. For weight management trials (example STEP programme): adults ≥18 years, BMI ≥30 or ≥27 with at least one weight-related comorbidity, history of at least one self-reported unsuccessful dietary effort. Overly restrictive criteria improve internal validity (cleaner data) but reduce generalisability (may not reflect real-world patient populations). Recent FDA guidance encourages broadening inclusion criteria to improve demographic representation.

Consent documents must include: study purpose, procedures, duration, experimental aspects, risks and discomforts, potential benefits, alternatives to participation, confidentiality protections, compensation for injury, contacts for questions, and statement of voluntary participation with right to withdraw. Consent must be obtained before any study procedures and re-obtained if the protocol changes materially. Special considerations: children require assent (age-appropriate consent) plus parental permission; patients with cognitive impairment may require legally authorised representative consent; emergency research has specific waiver provisions. For peptide drug trials, consent must specifically address injection-related risks, immunogenicity potential, and any class-specific safety concerns (e.g. thyroid C-cell tumour risk for GLP-1 RAs).

IRBs are composed of at least 5 members including scientists, non-scientists, and community representatives not affiliated with the institution. They evaluate: scientific design (is the study designed to answer the stated question?), risk-benefit ratio (do potential benefits justify the risks?), participant selection (is recruitment fair and non-coercive?), informed consent (is the document complete and comprehensible?), privacy and data protection, and vulnerable population protections. IRBs conduct initial review (before the study starts), continuing review (at least annually), and review of amendments and adverse event reports. Central IRBs can review multi-site trials, reducing duplication. In the US, IRBs are regulated under 45 CFR 46 (Common Rule) and 21 CFR 56 (FDA regulations).

ITT analysis follows the principle that the groups should remain comparable as randomised. Including all randomised patients (even dropouts) prevents: selection bias (analysing only 'good' patients), attrition bias (differential dropout between groups), and inflated effect estimates. Handling missing data in ITT: multiple imputation, mixed-effects models for repeated measures (MMRM), or conservative assumptions (e.g. baseline observation carried forward for primary outcomes). For weight management trials, a key ITT distinction is the treatment policy estimand (what is the effect of being assigned to treatment, including those who discontinue?) vs the trial product estimand (what is the effect in patients who continue treatment?). Both are reported but the treatment policy estimand is the primary regulatory analysis.

Interim analyses are performed by an independent statistical centre with unblinded data access, reviewed by the DSMB. Statistical approaches to maintain overall Type I error include: O'Brien-Fleming boundaries (very conservative early, allowing significance at final analysis close to conventional p<0.05), Lan-DeMets alpha-spending function (flexible timing of analyses), and Haybittle-Peto rule (requiring p<0.001 at interim, preserving p<0.05 at final analysis). For cardiovascular outcomes trials (CVOTs), interim analyses may occur after 50% and 75% of target events have accumulated. The SELECT trial's interim analysis showed such strong benefit that the DSMB recommended stopping the trial early — semaglutide 2.4mg met its primary cardiovascular endpoint before the planned final analysis date.

ClinicalTrials.gov was established in 2000 under the FDA Modernization Act of 1997. Registration became mandatory for most clinical trials involving FDA-regulated products under the FDA Amendments Act of 2007 (FDAAA 801). The International Committee of Medical Journal Editors (ICMJE) requires trial registration as a condition of publication. Each NCT record contains: study identification, design information, eligibility criteria, outcome measures, sponsor/collaborator, recruitment status, results (when available), and responsible party contacts. PeptideTrace's ClinicalTrials.gov sync pipeline queries the API v2 across all 185 compounds every 23 hours, tracking 13,223 trials across 132 compounds in steady state. The NCT number is the permanent link between PeptideTrace trial data and the full ClinicalTrials.gov record.

Non-inferiority margin (delta) definition requires clinical judgement: the margin represents the largest treatment difference that would still be considered clinically acceptable. Margin selection considers: historical effect of the active comparator vs placebo (the margin must be smaller than the proven effect to ensure the new drug is better than placebo — assay sensitivity), clinical importance, and regulatory guidance. Statistical test: one-sided 95% CI of the treatment difference must have its lower bound above -delta. Example: if semaglutide reduces HbA1c by 1.5% and the non-inferiority margin is 0.3%, the new drug must show HbA1c reduction with lower 95% CI bound >1.2%. Both ITT and PP analyses are important — for non-inferiority, PP is the primary analysis because ITT bias favours non-inferiority.

NNT = 1 / absolute risk reduction. Example: if a GLP-1 RA reduces 3-year MACE rate from 8.0% (placebo) to 6.4% (treatment), ARR = 1.6%, NNT = 1/0.016 = 63. This means 63 patients must be treated for 3 years to prevent one MACE event. NNT is time-dependent and condition-specific. Lower NNTs indicate greater treatment benefit. NNT for achieving ≥5% weight loss with semaglutide 2.4mg is approximately 1.3-1.5 (almost every treated patient achieves this threshold), reflecting the large treatment effect. NNT should be balanced against number needed to harm (NNH) for key adverse events to assess overall benefit-risk.

The three main observational designs — cohort, case-control, and cross-sectional — have complementary strengths. Cross-sectional studies examine exposure and outcome at a single time point (useful for prevalence estimation but cannot establish temporal sequence). All observational studies face confounding: in GLP-1 RA observational research, patients prescribed GLP-1 RAs may differ systematically from those on other treatments (confounding by indication). Statistical methods to address confounding include: multivariable regression, propensity score matching (balancing measured confounders), instrumental variable analysis (exploiting natural quasi-randomisation), and difference-in-differences analysis (comparing pre/post trends). Regulatory agencies increasingly use real-world evidence from observational studies to supplement clinical trial data.

OLE studies are offered to participants who complete the randomised phase, allowing those on placebo to cross over to active treatment and those on active treatment to continue. This design provides: long-term safety data (rare events may only emerge after years of exposure), durability of efficacy, immunogenicity data over extended periods, and real-world-like data on adherence and dose adjustments. Key limitations: no concurrent control group (making it difficult to attribute changes to the drug vs natural disease progression), selection bias (participants who tolerated and benefited from treatment are more likely to continue), and missing data from dropouts. OLE results for semaglutide showed sustained ~15% weight loss over 2 years of continued treatment.

Open-label designs are used when: blinding is technically impossible (e.g. comparing injectable vs oral formulations), the study is primarily assessing safety rather than efficacy, in compassionate use settings, or for Phase I dose-finding where investigators need to observe real-time effects. Open-label extension studies are the most common application in peptide drug development — participants who completed a blinded trial continue on active treatment for 1-5+ years, providing long-term safety data (adverse events, immunogenicity, laboratory parameters) and durability of efficacy data. Open-label extension data for GLP-1 RAs have demonstrated sustained weight loss and glycaemic control over 2-4+ years of continuous treatment.

P-values are calculated from the test statistic (t-test, chi-square, log-rank test, etc.) and the assumed null distribution. Common misconceptions: p<0.05 does NOT mean the probability that the result is true is 95%, nor that there is a 5% probability the null hypothesis is true. The p-value is conditional on the null hypothesis being true — it is the probability of observing data at least as extreme as what was observed, assuming no true difference exists. Bayesian approaches (calculating posterior probabilities of the treatment effect given the data) provide a more intuitive probability statement but require specifying prior beliefs. P-values are influenced by sample size — very large trials can produce small p-values for clinically trivial differences. Always evaluate p-values alongside effect size and confidence intervals.

PRO instruments must be validated (demonstrating reliability, validity, and responsiveness to change) for the specific population and condition. FDA PRO guidance (2009) established standards for PRO instrument development and qualification. Common PROs in peptide drug trials: IWQOL-Lite-CT (Impact of Weight on Quality of Life — Clinical Trials version, for obesity), SF-36 (general health status), EQ-5D (health utility for cost-effectiveness), WPAI (Work Productivity and Activity Impairment), and condition-specific instruments (e.g. FACT for cancer). PRO data capture increasingly uses electronic systems (ePRO) via smartphones or tablets, improving data quality and reducing missing data compared to paper questionnaires. PRO data can support label claims if based on pre-specified, well-powered analyses.

The PP population typically excludes: participants with major protocol violations (wrong drug, wrong dose), participants who discontinued treatment before a minimum exposure duration, participants who took prohibited concomitant medications, and participants with inadequate compliance (<80% of prescribed doses). PP results typically show larger treatment effects because the compliant, tolerant subpopulation benefits most. The difference between ITT and PP results (the ITT-PP gap) is clinically informative: a large gap suggests issues with tolerability or adherence. For non-inferiority trials, PP analysis is considered the primary analysis (because ITT can bias toward finding non-inferiority by including non-compliant patients in both groups).

Phase I designs include: single ascending dose (SAD) — small groups receive increasing single doses with safety monitoring between cohorts; multiple ascending dose (MAD) — groups receive repeated doses at each level to assess accumulation and steady-state PK; and food effect studies (assessing whether food alters absorption). First-in-human (FIH) starting dose is typically calculated from preclinical NOAEL using allometric scaling and a safety factor (usually 1/10th of the human equivalent dose of the animal NOAEL). For peptide drugs, Phase I establishes: maximum tolerated dose, dose-limiting toxicities, pharmacokinetic profile (Cmax, AUC, t1/2, CL, Vd), preliminary pharmacodynamic markers, and immunogenicity. Healthy volunteer studies are typical except for cytotoxic peptides (e.g. proteasome inhibitors) where Phase I enrols patients.

Phase IIa (proof of concept) typically enrols 20-100 patients to determine whether the drug produces the expected pharmacological effect in the target disease. Phase IIb (dose-ranging) enrols 100-300 patients across multiple dose groups to identify the optimal dose(s) for Phase III. For GLP-1 RAs, Phase II endpoints typically include HbA1c change, fasting plasma glucose, body weight change, and tolerability at each dose level. The dose selection decision between Phase II and Phase III is critical — choosing too low a dose risks insufficient efficacy, while too high risks excessive side effects and trial failure. Adaptive Phase II designs can modify dose groups based on interim results, improving efficiency.

Phase III characteristics: randomised, controlled (placebo and/or active comparator), double-blind, multicentre (often multinational), adequately powered (typically 80-90% power), pre-specified primary and secondary endpoints, intention-to-treat primary analysis, and pre-registered protocol. Regulatory agencies typically require two adequate and well-controlled pivotal trials (or one very large trial with robust results). GLP-1 RA Phase III programmes have been among the largest in metabolic medicine: the SUSTAIN programme (semaglutide for diabetes, 8 Phase III trials), STEP programme (semaglutide for weight, 5 Phase III trials), SURPASS programme (tirzepatide for diabetes, 5 Phase III trials), and SURMOUNT programme (tirzepatide for weight, 4 Phase III trials).

Phase IV studies serve multiple purposes: detecting rare adverse events (those occurring in <1:1,000-10,000 patients may not be captured in Phase III samples); evaluating long-term safety over years of use; comparing effectiveness against other treatments in real-world practice; studying use in special populations (elderly, paediatric, renal/hepatic impairment, pregnancy); and investigating new potential indications. Regulatory authorities can mandate Phase IV studies as post-marketing requirements (PMRs) or commitments (PMCs). For GLP-1 RAs, cardiovascular outcomes trials (CVOTs) serve as both Phase IV safety studies and efficacy trials for cardiovascular benefit. The SELECT trial (semaglutide) demonstrated cardiovascular benefit in patients with obesity without diabetes, expanding the evidence base beyond glucose control.

Pivotal trials are designed in close consultation with regulatory agencies (pre-IND meetings with FDA, scientific advice from EMA). Key pre-specifications include: primary endpoint, statistical hypothesis, sample size justification, analysis population, handling of missing data, multiplicity adjustment, and subgroup analyses. Results must demonstrate both statistical significance and clinical meaningfulness. For new indications of approved drugs (e.g. semaglutide for weight management), a new pivotal programme is required even though the drug is already approved for another indication. Pivotal trial data form the core of regulatory submissions (NDA/BLA/MAA) and are the primary basis for the benefit-risk assessment that determines approval.

Placebo-controlled designs require ethical justification — they are appropriate when no proven effective treatment exists for the condition, when withholding treatment poses no serious risk, or when adding the new drug to standard of care (placebo + SoC vs drug + SoC). For GLP-1 RA diabetes trials, placebo controls are justified as add-on to existing therapy (metformin, sulfonylureas, or insulin). The placebo response in weight management trials (typically 2-4% weight loss from lifestyle intervention alone) provides important context for interpreting drug effects. Nocebo effects (negative expectations causing adverse events in the placebo group) also occur — in GLP-1 RA trials, nausea rates of 4-10% in placebo groups illustrate this phenomenon.

Platform trials use a master protocol and shared infrastructure to evaluate multiple treatments simultaneously. Advantages: shared control group (reducing total sample size and placebo exposure), continuous operation (adding new arms as they become available, dropping arms for futility), and standardised data collection. RECOVERY (COVID-19) and I-SPY (breast cancer) are prominent platform trial examples. For peptide therapeutics, a metabolic platform trial could compare multiple GLP-1 RAs, dual agonists, and triple agonists against a shared placebo or standard-of-care control, with the ability to add emerging agents over time. Platform trials require complex governance structures to manage competing sponsor interests.

The preclinical programme for a peptide drug typically includes: in vitro receptor binding and functional assays (determining potency, selectivity, and mechanism), cell-based assays (confirming biological activity), ADME studies (stability in plasma, microsomal stability, permeability), pharmacology studies in relevant animal models (dose-response, time course, comparison to standard of care), safety pharmacology (cardiovascular/hERG, respiratory, CNS safety), toxicology (acute, subacute, chronic toxicity in two species — typically rodent and non-rodent), genotoxicity (Ames test, chromosomal aberration, micronucleus), and reproductive toxicity (fertility, embryo-foetal development, peri/postnatal development). For peptide drugs targeting human-specific receptors, transgenic animals expressing the human receptor may be needed for relevant pharmacology studies.

The primary endpoint determines the trial's sample size calculation, statistical analysis plan, and regulatory conclusion. It must be pre-specified, clinically meaningful, measurable, and achievable within the trial duration. For diabetes trials: change in HbA1c from baseline at 24-52 weeks. For weight management trials: percentage change in body weight from baseline AND proportion achieving ≥5% weight loss (co-primary endpoints). For cardiovascular outcomes trials: time to first MACE (3-point: CV death, non-fatal MI, non-fatal stroke). For cancer trials: overall survival or progression-free survival. The FDA and EMA may have different preferred primary endpoints for the same condition, requiring sponsors to pre-plan analyses acceptable to both agencies.

Phase IIa proof-of-concept studies typically enrol 20-100 patients and assess whether the drug produces detectable biological or clinical effects over 4-12 weeks. They may use biomarker endpoints (faster to measure than clinical outcomes) and enriched populations (selecting patients most likely to respond based on disease severity, genetic markers, or biomarker levels). For a new GLP-1 RA, PoC might demonstrate: HbA1c reduction ≥0.5% vs placebo at 12 weeks. For a research peptide transitioning from preclinical to clinical, PoC demonstrates that the mechanism of action identified in animal models translates to human biology. Failed PoC saves enormous development costs by identifying ineffective compounds before Phase III investment (~$100-300M for a typical Phase III programme).

Common reasons for amendments: updated safety information (requiring new exclusion criteria or monitoring), emerging data suggesting dose modification, slower-than-expected recruitment (broadening eligibility criteria), regulatory agency request (modified endpoint or control group), and practical operational issues. Substantial amendments (affecting safety, design, or scientific quality) require IRB/REC and regulatory approval before implementation — this process can take 4-12 weeks. Non-substantial amendments (administrative changes) may require only notification. Amendments affect all subsequently enrolled participants; already-enrolled participants may need re-consent if the changes materially affect their risk-benefit assessment. Frequent amendments can indicate poor initial protocol design.

Deviations are categorised as: minor (not expected to affect participant safety or data integrity — e.g. visit window slightly exceeded), major (may affect safety or data integrity — e.g. wrong dose administered, prohibited medication used), and critical (directly compromises participant safety — e.g. enrolled participant who met exclusion criteria for a safety reason). Major and critical deviations must be reported to the IRB/REC and sponsor. Systematic deviations (patterns affecting multiple sites) may indicate training deficiencies or protocol design problems. In regulatory submissions, deviation rates and types are reviewed as indicators of trial quality. Participants with critical deviations may be excluded from the PP analysis population.

Health-related quality of life (HRQoL) instruments capture physical functioning (mobility, self-care), psychological wellbeing (anxiety, depression, body image), social functioning (relationships, work), and symptom burden. The IWQOL-Lite for weight management includes physical function, self-esteem, sexual life, public distress, and work dimensions. In GLP-1 RA weight management trials, significant improvements in HRQoL have been demonstrated alongside weight loss, supporting the argument that weight loss with these drugs improves overall wellbeing beyond metabolic parameters. Health utility measures (EQ-5D) can be used in health economic analyses to calculate quality-adjusted life years (QALYs) for cost-effectiveness assessment — critical for reimbursement decisions by bodies like NICE (UK) and ICER (US).

Modern trials use computer-generated randomisation through interactive response technology (IRT/IXRS) systems that assign treatment in real-time when a participant is enrolled. Stratified randomisation ensures balance for key prognostic factors: common stratification variables include baseline HbA1c (for diabetes trials), baseline BMI (for weight trials), geographic region, and background therapy. Block randomisation (blocks of 4-8) ensures approximately equal allocation within each stratum. Allocation concealment (the process of preventing foreknowledge of treatment assignment) is critical — sequentially numbered, sealed envelopes are the minimum acceptable method, but centralised IRT systems are the gold standard. Broken allocation concealment is the single most important source of bias in clinical trials.

Randomisation methods include: simple randomisation (coin-flip equivalent — may produce unequal groups by chance), block randomisation (ensures equal group sizes within sequential blocks), stratified randomisation (balances key prognostic factors across groups), and adaptive/response-adaptive randomisation (adjusts allocation based on interim data). Allocation concealment ensures that the person enrolling participants cannot predict the next assignment, preventing selection bias. Control types: placebo (identical inactive treatment), active comparator (existing therapy — e.g. semaglutide Phase III trials compared to sitagliptin, empagliflozin, and insulin), and standard of care (whatever treatment the physician would normally prescribe). The choice of control affects the trial's ability to demonstrate superiority, non-inferiority, or equivalence.

RWE sources include: electronic health records (longitudinal patient data), insurance claims databases (treatment patterns, outcomes, costs), patient registries (disease-specific or drug-specific), personal health records and wearables, and social media/patient forums. Regulatory use: FDA Framework for RWE (2018) outlines how RWD can support new indications, label changes, and post-marketing requirements. The 21st Century Cures Act encourages FDA to use RWE. For peptide drugs, RWE has been used to: assess real-world GLP-1 RA effectiveness and adherence (often lower than trial efficacy), investigate post-marketing safety signals, compare treatments not directly compared in head-to-head trials (indirect treatment comparisons), and study outcomes in populations underrepresented in trials (elderly, paediatric, diverse ethnicities).

Registration trials are designed specifically to support regulatory submission. This means: the primary endpoint must be accepted by the target regulatory agency, the control group must be appropriate (agency preferences may differ — FDA may prefer placebo-controlled while EMA may prefer active comparator), the patient population must reflect the intended indication, the trial duration must be sufficient, and the statistical analysis plan must meet regulatory standards. The Special Protocol Assessment (SPA) process with the FDA provides written agreement on the design before the trial begins, ensuring that if the trial succeeds as designed, the results will support approval. Registration trials for peptide drugs typically cost $50-300M and take 2-4 years to complete.

RRR = (control event rate - treatment event rate) / control event rate × 100%. RRR is independent of baseline risk but can be misleading without absolute context. Example: RRR of 50% sounds impressive, but if the event rate decreases from 2% to 1% (ARR = 1%, NNT = 100), the clinical impact is modest. Conversely, RRR of 20% with baseline rate of 40% (ARR = 8%, NNT = 13) is clinically impactful. For GLP-1 RA cardiovascular trials, the ~20% MACE RRR translates to absolute risk reductions that depend on baseline cardiovascular risk of the population studied. Reporting both RRR and ARR (with NNT) provides the most complete picture of treatment benefit.

UK RECs operate under the Health Research Authority (HRA) and review all research involving NHS patients, tissue, or data. The National Research Ethics Service (NRES) coordinates REC review across the UK. REC composition includes expert and lay members. Review considers: scientific merit, participant safety, informed consent, confidentiality, and community benefit. EU RECs operate under the Clinical Trials Regulation (536/2014) with a coordinated assessment procedure for multinational trials. Key differences from US IRBs include: RECs tend to include more lay members, have standardised operating procedures, and follow specific legal frameworks (EU CTR, UK statutory instruments) rather than the more institution-based US approach.

Run-in period purposes: establish stable baseline measurements (reducing variability), identify and exclude non-compliant participants (improving adherence in the randomised phase), wash out effects of prior medications, and standardise background therapy. Placebo run-in (all participants receive placebo) identifies strong placebo responders who can be excluded, enriching the randomised population for drug-responsive patients. Active run-in (all participants receive active treatment) identifies intolerant patients before randomisation, reducing post-randomisation dropout. Run-in periods can bias trials — enrichment designs may overestimate the drug's benefit in the general population by excluding non-responders and non-tolerators.

Secondary endpoints are hierarchically tested (using a gatekeeping strategy) to control the overall false positive rate. Only if the primary endpoint is statistically significant can secondary endpoints be formally tested in a pre-specified order. If any secondary endpoint fails, all subsequent endpoints in the hierarchy are considered exploratory (not confirmatory). In GLP-1 RA weight management trials, the hierarchical secondary endpoint order might be: ≥10% weight loss proportion → ≥15% weight loss proportion → waist circumference change → systolic BP change → lipid changes → PRO measures. This hierarchy is pre-registered and reflects clinical importance priorities. Additional endpoints tested outside the hierarchy are considered exploratory regardless of their p-values.

SAE criteria (ICH E2A definition): results in death, is life-threatening (immediate risk of death at time of event), requires inpatient hospitalisation or prolongation of existing hospitalisation, results in persistent or significant disability/incapacity, is a congenital anomaly/birth defect, or is a medically important event that may jeopardise the patient or require intervention. SAEs must be reported to the sponsor within 24 hours. Suspected unexpected serious adverse reactions (SUSARs) — serious reactions not listed in the investigator's brochure — must be reported to regulatory authorities within 7 (fatal/life-threatening) or 15 (other serious) calendar days. SAE rates are a key safety metric in regulatory submissions and product labelling.

Single-blind (participant-blind) designs are used when investigator blinding is impractical — for example, when the treatments require different administration procedures that the investigator must perform (different injection techniques, different monitoring requirements). Participant blinding still reduces the placebo effect, reporting bias, and behavioural changes that knowledge of treatment assignment might cause. Single-blind designs are acceptable for some regulatory submissions but are considered weaker evidence than double-blind trials. Blinded outcome assessment (PROBE design — Prospective Randomised Open Blinded Endpoint) can partially compensate for lack of investigator blinding by ensuring outcomes are evaluated by individuals unaware of treatment assignment.

The conventional p<0.05 threshold means accepting a 5% false positive rate (Type I error). For multiple comparisons (testing many endpoints), adjusted thresholds are needed: Bonferroni correction (divide alpha by number of tests — very conservative), Holm-Bonferroni (step-down procedure), or hierarchical testing (gate-keeping, where secondary endpoints are only tested if the primary succeeds). Effect size matters alongside significance: Cohen's d (standardised mean difference) provides context: 0.2 = small, 0.5 = medium, 0.8 = large effect. For GLP-1 RA weight management trials, effect sizes for weight loss are typically very large (Cohen's d >1.0), meaning results are both statistically and clinically significant.

Stratification variables must be: prognostic (genuinely affecting the outcome), measurable at baseline (before randomisation), and limited in number (too many strata with small blocks leads to imbalance). Typical stratification for GLP-1 RA trials: baseline HbA1c (<8.5% vs ≥8.5%), background OAD (metformin alone vs metformin + sulfonylurea), and geographic region. For weight management trials: baseline BMI (30-<35 vs 35-<40 vs ≥40) and presence of type 2 diabetes. Stratified analysis ensures treatment effect estimates are adjusted for these important baseline variables. Minimisation (a dynamic allocation method) can balance many more variables than stratified block randomisation but is more complex to implement.

Superiority testing uses a two-sided hypothesis: H0 (null): treatment effect = 0 (no difference); H1 (alternative): treatment effect ≠ 0. The trial is powered (typically 80-90%) to detect a pre-specified minimum clinically important difference (MCID) at a significance level of α = 0.05 (two-sided). Sample size calculation depends on: expected effect size, variability (standard deviation) of the primary endpoint, desired power, significance level, and expected dropout rate. For weight management trials, semaglutide Phase III programmes were powered to detect superiority over placebo with co-primary endpoints (% weight change AND ≥5% weight loss proportion) — both must be statistically significant for the trial to succeed.

Surrogate endpoint validation requires demonstrating that the surrogate lies on the causal pathway of the disease, that changes in the surrogate reliably predict changes in the clinical outcome, and that treatment effects on the surrogate predict treatment effects on the outcome. HbA1c is a validated surrogate for diabetes complications based on decades of outcome data (DCCT, UKPDS trials). Body weight change as a surrogate for cardiovascular outcomes is becoming better validated (SELECT trial provided evidence that weight loss with semaglutide reduces cardiovascular events). The FDA's accelerated approval pathway allows approval based on surrogate endpoints reasonably likely to predict benefit, with confirmatory trials required post-approval.

TEAE definition: any AE with onset on or after first dose of study medication, or any pre-existing condition that worsens in severity or frequency after first dose. TEAEs are distinguished from pre-treatment AEs (events occurring before first dose, captured during screening/run-in) and post-treatment AEs (events occurring after the last dose + a defined follow-up period, typically 5 half-lives). TEAE tables in clinical study reports show: overall incidence, incidence by System Organ Class and preferred term, severity breakdown, relationship to study drug assessment, and events leading to discontinuation. For regulatory review, TEAE data are compared between treatment and control groups using exposure-adjusted incidence rates to account for different treatment durations.

Registration requirements: FDAAA 801 mandates registration of interventional trials of FDA-regulated drugs/biologics/devices within 21 days of first patient enrolment. WHO/ICMJE requires registration before first patient enrolment as a condition of journal publication. Information registered includes: descriptive information (title, sponsor, conditions, interventions), recruitment information (eligibility criteria, enrollment target, locations), design information (study type, phase, masking, allocation, endpoints), and administrative information (IND/IDE number, responsible party). Results reporting requirements mandate submission of summary results within 12 months of primary completion. Non-compliance with registration/reporting requirements can result in civil penalties up to $10,000/day and withholding of NIH funding.

Emergency unblinding procedures: the investigator contacts the sponsor's 24/7 safety hotline (or accesses the IRT system) and receives the treatment assignment only if knowledge of the assignment would change the patient's acute medical management. The event is documented with justification. The unblinded patient typically continues in the study (to preserve ITT analysis integrity) unless the investigator determines discontinuation is necessary. Planned unblinding occurs at study completion or after database lock. The DSMB has access to unblinded data throughout the trial for safety monitoring but operates under strict confidentiality. Any accidental unblinding (e.g. through laboratory values, distinctive side effects, or labelling errors) is documented as a protocol deviation.

Washout duration = 5-7 × drug half-life (to reduce residual drug concentration to <3% of steady-state levels). For weekly semaglutide (t1/2 ~1 week): washout would require 5-7 weeks for pharmacokinetic clearance, though pharmacodynamic effects (appetite suppression, GI adaptation) may persist longer. Carryover effects (persistent drug effects after washout) complicate crossover trial interpretation and are assessed statistically. In practice, washout periods must balance completeness of drug elimination against participant burden and disease management needs. For drugs with irreversible effects (e.g. GnRH agonist-induced receptor downregulation), the washout must be long enough for receptor recovery and hormonal normalisation, which may take weeks beyond drug elimination.