{
  "id": "clinical-trial",
  "title": "Clinical Trial",
  "lang": "en",
  "source_lang": "en",
  "summary": "A prospective study assigning people to health interventions to evaluate their effects on health outcomes.",
  "category": "health",
  "subcategory": "medicine",
  "type": "concept",
  "aliases": [
    "Interventional trial",
    "Clinical interventional study"
  ],
  "generated_at": "2026-10-15",
  "model": "gpt-6.1-sol",
  "url": "https://aiwiki.page/en/clinical-trial/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/clinical-trial/",
      "json": "https://aiwiki.page/zh/clinical-trial.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A clinical trial is a research study that prospectively assigns human participants, individually or in groups, to one or more health-related interventions to evaluate their effects on health outcomes. Interventions include medicines, vaccines, medical devices, surgical procedures, behavioral treatments, and changes in healthcare delivery. Clinical trials are distinguished from observational studies by the investigator’s assignment of interventions; they need not be randomized or include a control group. (who.int)\n\nClinical trials provide evidence about the benefits and harms of interventions and inform medical-product assessment, clinical practice, and public health. Their reliability depends on clearly defined questions, appropriate experimental design, systematic data collection, and suitable statistical analysis. The randomized controlled trial is an important design, but clinical research also uses nonrandomized and specialized approaches suited to different questions and stages of development. (who.int)\n\nThe conduct of clinical trials combines scientific methods with protections for participants, including independent ethical review and informed consent. Protocols, safety oversight, registration, and transparent reporting connect the planning of a study with the interpretation and use of its findings. Clinical-trial practice also encompasses participant representation, international research coordination, and designs that incorporate adaptive methods or decentralized activities. (fda.gov)"
    },
    {
      "heading": "Definition, Scope, and Research Purposes",
      "text": "A clinical trial is an interventional study in which human participants, individually or in groups, are prospectively assigned to a health-related intervention so that its effects on health outcomes can be evaluated. The World Health Organization uses this definition for trial registration. “Prospective assignment” means that a research plan specifies how participants will receive an intervention before its effects are assessed; merely following people forward in time does not make a study a trial. Assignment need not involve randomization, and a trial need not have a comparison group: single-arm studies assessing safety, drug disposition, or tolerability can qualify. Participants may be healthy volunteers rather than people with an established disease. (who.int)\n\nClinical trials are a subset of clinical research, which also encompasses epidemiological, behavioral, and health-services investigations. Their defining distinction from an observational study is investigator-directed assignment to the intervention being evaluated. For example, comparing outcomes among patients whose physicians independently prescribed different medicines is observational; assigning those medicines through a research protocol is interventional. Observational studies can have protocols, scheduled assessments, and prospective follow-up without becoming trials. Routine care primarily addresses an individual’s clinical needs, whereas a trial systematically evaluates a research question, even when both use the same approved treatment. Secondary analysis of existing health records or stored specimens alone is not a clinical trial. (nih.gov)\n\nInterventions extend beyond experimental medicines. They include drugs; biological products such as vaccines and cell therapies; medical devices; surgical and radiological procedures; and behavioral approaches such as exercise, dietary change, and cognitive therapy. Trials may also assign ways of delivering care, including telemedicine, or changes in clinical processes. The intervention can therefore be a product, a procedure, a behavioral program, or an organizational strategy. Neither novelty nor lack of marketing authorization is essential: trials can investigate established treatments, alternative combinations, or different uses of an existing intervention. (who.int)\n\nResearch purposes distinguish several overlapping categories. Treatment trials evaluate interventions intended to control disease or improve symptoms. Prevention trials investigate ways to reduce disease occurrence, including vaccination or behavioral change. Screening trials assess strategies for detecting disease before symptoms appear, whereas diagnostic trials investigate approaches to identifying or characterizing suspected disease, including imaging strategies. Supportive-care trials evaluate ways to reduce symptoms, treatment-related burdens, or functional limitations and improve quality of life. Health-service improvement trials test changes in care delivery, organization, or management; they meet the interventional definition when care strategies are prospectively assigned and their effects on health outcomes are evaluated. These purposes do not correspond one-to-one with intervention types: a medicine may be preventive or therapeutic, and a behavioral program may support either aim. (cancer.gov)\n\nThe National Lung Screening Trial illustrates why evaluating a screening strategy involves more than demonstrating that a test detects abnormalities. Launched in 2002, it assigned 53,454 current or former heavy smokers aged 55–74 to three annual screening examinations using either low-dose computed tomography or chest radiography. Its primary results, published online on June 29, 2011, showed approximately 15–20% lower lung-cancer mortality in the computed-tomography group. Both groups received diagnostic technology, but the research purpose was screening, and the principal health question concerned deaths from lung cancer, not simply the number of tumors detected. (cancer.gov)"
    },
    {
      "heading": "Historical Development",
      "text": "On 20 May 1747, the Scottish naval surgeon James Lind began an early controlled comparison of scurvy treatments aboard HMS Salisbury. His account, published in A Treatise of the Scurvy in 1753, described 12 affected sailors divided into six pairs, sharing accommodation and a common basic diet but receiving different remedies: cider, dilute sulfuric acid, vinegar, seawater, oranges and lemons, or a medicinal paste. The citrus supply lasted six days; by then, one recipient was fit for duty and the other sufficiently recovered to nurse the remaining sick. Comparing treatments concurrently under similar conditions distinguished the experiment from an isolated report of recovery. It was not, however, a modern randomized controlled trial, nor should it be treated as the uncontested first clinical trial. (jameslindlibrary.org)\n\nThe British Medical Research Council’s streptomycin investigation, published on 30 October 1948, became a methodological landmark through its explicit protection against foreknowledge of treatment assignments. In patients with pulmonary tuberculosis, it compared streptomycin plus bed rest with bed rest alone; the reported analysis included 55 patients in the streptomycin group and 52 controls. Austin Bradford Hill prepared allocation sequences using random sampling numbers separately for each sex at each participating centre. Numbered, sealed envelopes were held centrally, and the appropriate envelope was opened only after a patient had been accepted for the trial. Investigators therefore could not know the next assignment when deciding whether to enroll someone. This combined random allocation with allocation concealment, rather than relying on a predictable sequence such as alternation. (bmj.com)\n\nThe trial also separated treatment delivery from aspects of outcome assessment: specialists assessing chest radiographs were unaware of treatment allocation. At six months, 4 of 55 streptomycin recipients and 14 of 52 controls had died, approximately 7% and 27%, respectively. These findings provided a quantified comparison against a concurrent control group, rather than attributing improvement to treatment solely because it followed administration of the drug. The study’s importance lay in its documented combination of comparison, concealed random assignment, and independent assessment—not in an uncontested claim to have invented every component. (jameslindlibrary.org)\n\nFormal participant protections developed through a partly separate history. The Nuremberg Code of 1947 and the Declaration of Helsinki, first adopted in 1964, preceded major changes in United States oversight. Following publicized research abuses, the National Research Act became law on 12 July 1974. It established the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research and directed the Department of Health, Education, and Welfare to require institutional review boards for covered research. Protection thus acquired a more formal institutional structure rather than depending solely on an individual investigator’s judgment. (hhs.gov)\n\nThe commission’s Belmont Report, published on 18 April 1979, articulated respect for persons, beneficence, and justice as basic principles of research ethics. It connected these principles to informed consent, assessment of risks and benefits, and selection of participants. Respect for persons encompassed autonomy and protection for people with diminished autonomy; beneficence required attention to possible harms as well as benefits; justice concerned the distribution of research burdens and benefits. The report was an ethical framework rather than a clinical-trial protocol or a statute, and it subsequently informed revisions to United States human-subject protection regulations in 1981. (hhs.gov)"
    },
    {
      "heading": "Phases of Clinical Development",
      "text": "Phase 0 denotes optional exploratory human studies conducted before conventional dose-escalation trials. These involve very limited exposure and have no therapeutic or diagnostic intent; microdose studies are one example. Their purpose is to assess whether a candidate warrants further development, rather than establish treatment efficacy. The United States Food and Drug Administration’s January 2006 guidance on exploratory investigational new drug studies describes these investigations as early phase I studies and gives seven days as an example of limited dosing duration. ClinicalTrials.gov uses “Early Phase 1” for studies formerly classified as phase 0. (fda.gov)\n\nPhase I investigates initial safety, tolerability, dose, and human pharmacology. Pharmacokinetics describes drug absorption, distribution, metabolism, and excretion; pharmacodynamics concerns biological effects and the relationship between exposure and response. Single- and multiple-dose studies help characterize adverse reactions, accumulation, and dosing intervals. Participants may be healthy volunteers or patients, depending on the product and anticipated risks; early trials of anticancer drugs commonly enroll people with cancer. FDA educational material gives approximately 20–100 participants as a typical phase I range, not a mandatory sample size. The resulting evidence guides doses and schedules for subsequent investigation rather than demonstrating that a drug is universally safe. (database.ich.org)\n\nPhase II evaluates preliminary activity or efficacy in patients with the intended condition while continuing safety assessment. Studies commonly include up to several hundred participants. Dose-ranging comparisons investigate whether greater exposure improves response, increases toxicity, or both, and help select a regimen for confirmatory testing. Exploratory work can also refine the target population and identify appropriate outcomes or biomarkers. A favorable preliminary result supplies a rationale for further testing, not necessarily a confirmatory conclusion about clinical benefit. (fda.gov)\n\nPhase III generally follows preliminary evidence of efficacy and seeks more definitive evidence about effectiveness and the benefit–risk relationship in the intended population. A randomized controlled trial may compare the selected regimen with established treatment or, where appropriate, placebo. FDA describes phase III studies as commonly involving several hundred to several thousand participants. Their larger scale and longer exposure can reveal adverse effects missed earlier, although uncommon or delayed harms may remain unresolved. These studies contribute evidence for marketing authorization and prescribing information; the phase designation itself does not guarantee a positive result or regulatory approval. (database.ich.org)\n\nPhase IV trials take place after marketing authorization and investigate additional questions about risks, benefits, or optimal use. They may compare an authorized medicine with another treatment or examine longer-term outcomes. An interventional phase IV study assigns treatment prospectively according to a research protocol. It is therefore distinct from observational postmarketing surveillance, which may analyze routine prescribing records, patient registries, or adverse-event reports without assigning treatment. Both contribute postauthorization evidence, but the fact that data were collected after approval does not by itself make an investigation a phase IV trial. (clinicaltrials.gov)\n\nThe phases describe development objectives rather than an obligatory sequence of separate studies. Programs can overlap stages or combine them in phase I/II or phase II/III trials; pharmacokinetic questions may also be investigated late in development. This framework primarily concerns drugs and biological products and does not apply uniformly to medical devices, surgery, or behavioral interventions. ClinicalTrials.gov consequently permits “Not Applicable” for trials without drug-development phases, explicitly including device and behavioral studies. (database.ich.org)"
    },
    {
      "heading": "Research Questions and Trial Protocols",
      "text": "Before enrollment begins, a trial protocol translates the research question into a prospective plan for conducting and analyzing the study. Its rationale identifies the unresolved clinical problem, reviews relevant prior evidence, and explains why the intervention, comparator, and proposed population are appropriate. This evidence includes previous findings on benefits and harms, rather than only studies suggesting efficacy. Objectives distinguish what the trial primarily seeks to establish from secondary or exploratory questions; confirmatory objectives are linked to explicit hypotheses and a prespecified comparison. The protocol also defines eligibility, intervention delivery, assessments, follow-up duration, and the principal analytical approach. These decisions connect the clinical question to an executable experimental design. (fda.gov)\n\nOperational specifications make the intended comparison reproducible across sites. For a drug study, these include dose, route, frequency, treatment duration, and conditions for modifying or stopping treatment; behavioral interventions similarly require descriptions of their content and delivery. A participant schedule distinguishes screening, baseline measurements, intervention periods, and subsequent assessments, including acceptable visit windows. Treatment duration and observation duration need not be identical. Feasibility assessment examines whether eligible participants, trained staff, equipment, and facilities are available, and whether recruitment and follow-up can be completed as planned. The International Council for Harmonisation’s E8(R1) framework treats operational feasibility as part of study quality: a scientifically relevant question is not adequately addressed by a protocol that cannot be implemented. (fda.gov)\n\nThe protocol and statistical analysis plan (SAP) perform complementary functions. The protocol establishes the principal statistical features; the SAP supplies technical details needed to execute them, including analysis populations, variable derivation, statistical models, treatment comparisons, handling of missing observations, and additional analyses. Their definitions must remain aligned with the objectives and measurements actually collected. For example, if the question concerns change from baseline at a particular follow-up visit, the planned analysis must identify that change and time point, rather than leave analysts to choose among several measurements after examining results. ICH E9 permits the separate SAP to be completed after the protocol, but specifies that it should be finalized before treatment assignments are unblinded, with formal records of both dates. (ema.europa.eu)\n\nPrespecification does not make a protocol immutable. New scientific information or problems identified during blinded data review may justify revisions. Changes to principal features require a protocol amendment; technical refinements that leave those features unchanged may instead be recorded in an updated SAP. Version histories document what changed, when, and why, allowing reviewers to distinguish prospective decisions from analyses developed after results became known. ICH E9 also requires departures from the planned analysis to be identified and explained in the study report. Additional exploratory analyses may be informative, but are distinguished from the analyses originally intended to test the trial’s hypotheses. (ema.europa.eu)\n\nSPIRIT 2025, published on April 28, 2025, updates the 2013 Standard Protocol Items: Recommendations for Interventional Trials guidance. It provides 34 minimum reporting items for randomized-trial protocols and a diagram showing enrollment, interventions, and assessments. Its revisions include an open-science section and greater attention to harms, intervention descriptions, and patient and public involvement. SPIRIT concerns the completeness and transparency of what a protocol reports, not certification that its scientific choices are valid: a fully described study may still have an unsuitable comparator, inadequate measurements, or an unjustified analysis. Its minimum checklist focuses on parallel-group randomized trials; specialized designs may require additional protocol content. (jamanetwork.com)"
    },
    {
      "heading": "Participants, Recruitment, and Representation",
      "text": "The target population comprises the people to whom a trial’s research question applies; the enrolled population consists of those actually recruited. Inclusion criteria specify required characteristics, such as a confirmed diagnosis, disease stage, or previous treatment, while exclusion criteria identify circumstances incompatible with participation, such as a clinically important drug interaction. Screening checks these requirements through medical histories, examinations, laboratory tests, or record review. For example, the National Lung Screening Trial enrolled current or former smokers aged 55–74 years with at least 30 pack-years of smoking exposure; former smokers must have quit within the preceding 15 years. These criteria defined the population studied, rather than establishing eligibility for all subsequent screening programmes. (dtc.cancer.gov)\n\nRecruitment involves identifying, approaching, and enrolling potential participants through clinical referrals, record searches, advertisements, registries, or community organizations. Eligibility alone does not ensure access: travel distance, unpaid time away from work, childcare, inaccessible facilities, and language barriers can prevent otherwise eligible people from enrolling. Recruitment planning therefore includes site selection, translated materials, accessible communication, transport arrangements, and feasible appointment schedules. Recording how many people were approached, screened, eligible, and enrolled—and why others did not participate—helps distinguish restrictive eligibility from practical barriers or decisions to decline participation. (fda.gov)\n\nRetention concerns continued participation and outcome collection over the planned follow-up period. Clear visit schedules, reminders, updated contact information, and arrangements that reduce unnecessary travel can support retention. Stopping the assigned intervention does not necessarily mean withdrawing from follow-up: a participant may discontinue medication but permit subsequent assessments. Loss to follow-up occurs when required outcome information can no longer be obtained, and differs from a documented decision to withdraw. Missing outcomes can distort comparisons when their absence relates to treatment response or adverse effects; statistical methods for addressing them do not replace efforts to maintain appropriate, voluntary contact. (fda.gov)\n\nRepresentation includes age, sex, race and ethnicity, geography, disability, comorbidity, and other characteristics relevant to the intended patient population. Children and older adults may be excluded through age limits or requirements that disproportionately remove people with multiple conditions. The United States National Institutes of Health’s Inclusion Across the Lifespan policy, applicable to grant applications with due dates from January 25, 2019, requires inclusion across ages unless scientific or ethical reasons justify exclusion. For pregnant people, fetal risks and changes in drug disposition require specific consideration rather than simple extrapolation from nonpregnant adults. FDA’s April 2018 draft guidance proposed a balanced approach to inclusion during pregnancy, with attention to potential fetal risk; it was issued as draft, nonbinding guidance. (grants.nih.gov)\n\nPatient and community involvement differs from recruitment: people affected by a condition can help determine research priorities without becoming trial participants. Contributions can include identifying meaningful outcomes, reviewing information materials, assessing the acceptability of procedures, and identifying burdens imposed on participants and caregivers. The World Health Organization’s 2024 guidance places such engagement across research planning, delivery, and dissemination. Community consultation can reveal, for example, that repeated hospital visits or an invasive procedure would make an otherwise scientifically feasible study unacceptable to its intended population; it does not substitute for individual informed consent. (who.int)\n\nRecruitment settings and restrictions shape external validity, the applicability of findings beyond those enrolled. A tertiary referral centre may recruit a different clinical population from primary-care practices, while exclusions based on organ dysfunction or concurrent illness can leave important treatment users unrepresented. In oncology, a 2017 joint statement by the American Society of Clinical Oncology and Friends of Cancer Research proposed broadening eligibility concerning brain metastases, HIV infection, younger age, prior malignancies, and organ dysfunction. NCI subsequently incorporated expanded eligibility language into protocol development. These changes address the distinction between restrictions necessary for safety or the research question and inherited criteria that unnecessarily narrow participation. (iris.who.int)"
    },
    {
      "heading": "Comparators, Randomization, and Masking",
      "text": "A placebo control resembles the investigational treatment but lacks its active component, allowing a comparison that accounts for expectations, treatment rituals, and spontaneous changes in illness. Placebo does not necessarily mean absence of care: an add-on trial can compare a new drug plus standard therapy with placebo plus the same therapy. An active-treatment control receives another treatment, usually one with established efficacy; the comparison measures relative performance rather than benefit over no treatment. Similar outcomes do not by themselves establish that both treatments work. A standard-care control receives usual clinical management, which may include several treatments and vary between sites. A no-intervention control receives no study intervention, although ordinary care may continue; unlike a matched placebo, it generally cannot conceal treatment assignment or equalize treatment-related attention. These distinctions are addressed in ICH E10, adopted in 2000 and issued as FDA guidance in May 2001. (fda.gov)\n\nAn external control consists of patients outside the trial’s treatment group, drawn from earlier records, another study, or a concurrent population elsewhere. Its defining feature is how patients enter the comparison, not which treatment they receive. Without random assignment, differences in disease severity, eligibility, accompanying care, calendar time, or outcome measurement can produce confounding. For example, improved supportive care can make a new treatment appear beneficial when compared with older records. Matching and statistical adjustment address measured differences but cannot reliably eliminate unmeasured ones. External controls may be considered when a randomized comparison is infeasible, particularly for rare conditions with predictable natural histories, but their causal interpretation requires stronger assumptions. (fda.gov)\n\nIn a randomized controlled trial, randomization uses a chance-based sequence, commonly generated by computer, rather than clinician preference, alternation, or birth date. It prevents systematic treatment selection based on known or unknown prognostic characteristics, although chance imbalances remain possible. Blocking maintains the allocation ratio over successive groups of assignments: in an illustrative 1:1 design, each completed block of four contains two assignments to each treatment in random order. Stratification creates separate randomization schedules within categories such as study site or disease severity, improving balance on selected prognostic factors. Small, fixed blocks can make later assignments predictable; concealed, randomly varying block sizes reduce that risk. (fda.gov)\n\nAllocation concealment prevents recruiters and prospective participants from learning the next assignment before enrollment is irrevocable. Otherwise, even a genuinely random sequence can be subverted by delaying enrollment or selectively recruiting particular patients. Central web-based or telephone randomization can release an assignment only after eligibility and enrollment are confirmed. Sequentially numbered, opaque, sealed containers offer another mechanism when properly secured and implemented. Concealment concerns the period before and through assignment; masking concerns knowledge after assignment. An open-label trial can therefore retain concealed allocation despite participants and clinicians subsequently knowing the treatment. (pmc.ncbi.nlm.nih.gov)\n\nMasking, also called blinding, can apply separately to participants, treating clinicians, outcome assessors, and statistical analysts. Participant masking limits expectation-related differences in symptom reporting; clinician masking reduces differential accompanying care; assessor masking limits judgment influenced by treatment knowledge; analyst masking can restrict data-dependent analytical choices. Reports therefore specify who was masked rather than relying on ambiguous labels such as “double-blind.” Identical packaging and appearance can support masking, although recognizable adverse effects may reveal assignments. In a double-dummy comparison of unlike formulations, each group receives its assigned active treatment plus a placebo resembling the other treatment. (consort-spirit.org)\n\nMasking participants or practitioners may be impracticable when comparing surgery with medication or delivering an intervention requiring active participation. Randomization can nevertheless remain feasible, with masked independent outcome assessment used where possible. Conversely, random assignment to withhold established lifesaving treatment may be unacceptable; this limits comparator choices rather than making every randomized design impossible. When neither randomization nor masking is feasible, standardized accompanying care, common assessment schedules, and independent review can reduce particular biases, but do not recreate the protection against confounding supplied by random assignment. (pmc.ncbi.nlm.nih.gov)"
    },
    {
      "heading": "Trial Design Architectures",
      "text": "In a parallel-group design, participants are assigned to separate groups that receive different interventions over the same study period; comparisons are between participants. In a crossover design, participants receive interventions in sequence—for example, A then B or B then A—with the sequence assigned by randomization. Each participant serves as their own control, reducing variation attributable to stable differences between people. However, an earlier treatment may continue affecting outcomes during a later period: this is carryover. A washout interval allows effects to dissipate, but must reflect the duration of biological effects, not merely drug elimination. Crossover designs therefore suit stable conditions and reversible effects better than curative interventions or rapidly progressive disease. Comparisons of two formulations of the same medicine for bioequivalence are a common application. Analyses must also account for treatment period and within-person dependence. (fda.gov)\n\nA factorial design assigns combinations of interventions. A \\(2 \\times 2\\) trial has four groups: A alone, B alone, A plus B, and neither. If interaction is negligible, A’s main effect can be estimated by comparing both groups receiving A with both groups not receiving A, while estimating B’s effect analogously. This can answer two questions efficiently within one participant population. A statistical interaction occurs when the effect of A differs according to whether B is present, on the specified measurement scale. A trial powered to detect main effects may nevertheless have insufficient power to detect an interaction; a nonsignificant interaction test does not establish that treatments act independently. (fda.gov)\n\nIn a cluster-randomized trial, groups such as practices, schools, or communities, rather than individuals, are allocated to interventions. This accommodates group-level delivery and can reduce contamination between study arms. Outcomes within a cluster tend to resemble one another, so participant numbers alone overstate the amount of independent information. For a simple parallel cluster design with equal cluster size \\(m\\) and intracluster correlation \\(\\rho\\), the conventional design effect is \\(1+(m-1)\\rho\\). Thus, with 50 participants per cluster and \\(\\rho=0.02\\), the multiplier is 1.98: almost twice as many participants are needed as under otherwise comparable individual randomization. Analysis must preserve the clustered structure rather than treat observations as independent. (bmj.com)\n\nA stepped-wedge design typically begins with all clusters under control conditions, then introduces the intervention at successive time points according to randomized transition sequences, until all clusters receive it. Outcomes are measured across periods, using the same individuals, successive samples, or a mixture. Because intervention observations occur disproportionately later, improvements caused by changing clinical practice, seasonality, or other calendar-time trends can be mistaken for treatment effects. Design and analysis must therefore address both clustering and time; the simple parallel-cluster design-effect formula does not suffice. Karla Hemming and colleagues’ 2015 methodological account emphasized these issues in evaluating staged service-delivery changes. (bmj.com)\n\nExplanatory and pragmatic describe aims rather than mutually exclusive allocation structures. Explanatory trials investigate effects under tightly controlled conditions; pragmatic trials assess effects under conditions resembling routine care. The PRECIS-2 tool, published in 2015, assesses nine design domains—including eligibility, setting, intervention delivery, and follow-up—on scales from 1, very explanatory, to 5, very pragmatic. A trial may occupy different positions across domains, and either aim can be pursued through individual or cluster randomization. (bmj.com)\n\nSingle-arm trials assign all participants to one intervention without a concurrent control arm. They can characterize responses, but cannot by themselves separate treatment effects from natural history or other influences. (fda.gov) N-of-1 trials, by contrast, make repeated treatment comparisons within one participant, commonly through multiple crossover periods with randomized sequences. They address individual treatment response rather than primarily estimating a population-average effect. Like other crossover studies, they require attention to carryover and stability over time; the 2015 CENT reporting guideline specifically addresses these prospective multiple-crossover experiments. (bmj.com)"
    },
    {
      "heading": "Outcomes and Measurement",
      "text": "A primary outcome addresses the trial’s principal research objective; secondary outcomes support that objective or assess additional effects, while exploratory outcomes investigate signals and generate hypotheses for later testing. These designations describe an outcome’s role, not its measurement type: survival, symptoms, or a laboratory measurement may occupy any category. An endpoint requires more precision than a domain such as “mortality” or “pain”: its definition includes the variable, assessment method, time point or observation period, and rule for summarizing measurements. For example, death from any cause within 30 days differs from time to death over two years. Protocols specify assessment schedules, permissible visit windows, instruments, and scoring procedures before results are known. (fda.gov)\n\nClinical outcomes directly concern survival, symptoms, or functioning. All-cause mortality records death regardless of attributed cause; functional status may be assessed through standardized performance tasks or ability to undertake daily activities. A patient-reported outcome comes directly from the participant without interpretation by a clinician or another person. Examples include pain intensity, fatigue, and perceived physical limitations. Patient reporting is therefore a method of assessment, not an alternative to a clinical outcome. Health-related quality of life encompasses several domains, commonly including physical, emotional, and social functioning; a symptom scale alone does not necessarily measure this broader construct. The FDA’s December 2009 guidance distinguished these concepts when describing instruments used to support medical-product labeling claims. (fda.gov)\n\nA biomarker measures a biological characteristic rather than directly describing how a person feels, functions, or survives. Blood pressure, circulating substances, and imaging measurements can indicate disease activity or a biological response to treatment. A surrogate endpoint substitutes for a clinical outcome because treatment-induced changes in it are expected to predict clinical benefit. Association with prognosis is insufficient: patients with better biomarker values may fare better without an intervention that changes the marker necessarily improving their outcomes. Validation requires mechanistic and clinical evidence supporting that prediction in a specified context. Blood-pressure reduction, for example, is an accepted surrogate for reduced stroke risk in appropriate settings; such acceptance does not establish that every biomarker or every intervention acting on it predicts benefit. (fda.gov)\n\nA composite endpoint combines distinct events under a predefined rule—for example, time to the first cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke. Combining events can increase the number of observed endpoints, but the components may differ in frequency, severity, and responsiveness to treatment. A favorable composite result therefore does not establish a reduction in every component: a frequent, less consequential event may dominate the finding. Component-specific results help explain what changed. The FDA’s October 2022 guidance on multiple endpoints distinguishes these event composites from multicomponent assessments that combine several measurements within each participant into an overall rating. (fda.gov)\n\nMeasurement quality depends on validity, reliability, and responsiveness. Validity concerns whether an instrument captures its intended concept; reliability concerns consistency under stable conditions, including agreement between assessors; responsiveness concerns detection of change. Standardized instructions, assessor training, consistent recall periods, and appropriately evaluated translations support comparable measurements. For events requiring clinical judgment, blinded adjudication committees can apply prespecified definitions to clinical records: an FDA review published in 2021 describes cardiovascular endpoints evaluated centrally by an independent committee unaware of treatment assignment. For symptom scores, clinically meaningful within-person change can be investigated using external anchors, such as patients’ global assessments of improvement. Such a threshold is instrument- and population-dependent and is not automatically a threshold for interpreting a between-group mean difference. (fda.gov)"
    },
    {
      "heading": "Statistical Objectives and Sample Size",
      "text": "A superiority trial tests whether one intervention produces a better outcome than its comparator. Noninferiority instead asks whether any disadvantage is smaller than a prespecified, clinically acceptable margin; equivalence requires excluding unacceptable differences in both directions. If \\(\\theta\\) denotes the treatment difference, with positive values favoring the experimental intervention, noninferiority tests \\(H_0:\\theta\\leq-\\Delta\\), whereas equivalence seeks to establish \\(-\\Delta_L<\\theta<\\Delta_U\\). The corresponding confidence interval must exclude the relevant noninferiority boundary or lie wholly within the equivalence limits, using confidence levels matched to the testing procedure. Failure to demonstrate superiority does not establish either noninferiority or equivalence. These distinctions appear in the 1998 ICH E9 statistical guideline. (fda.gov)\n\nComparison margins must be justified before examining treatment results, rather than chosen to accommodate the available sample. For an active-controlled noninferiority trial, justification combines clinical judgment with historical evidence of the comparator’s benefit over placebo. The FDA’s November 2016 guidance distinguishes \\(M_1\\), a conservative estimate of the active control’s effect, from \\(M_2\\), the maximum clinically acceptable loss of that effect. The selected margin generally preserves a clinically important portion of the established benefit. Interpretation also depends on whether historical evidence remains applicable to the present population, treatment delivery, and outcome assessment; two similarly ineffective treatments can otherwise appear noninferior. (fda.gov)\n\nSample-size planning specifies an anticipated effect size, outcome variability or event rates, the significance level \\(\\alpha\\), and statistical power \\(1-\\beta\\): the probability of rejecting the null hypothesis under a specified alternative. Smaller target differences, greater variability, stricter significance thresholds, or higher desired power generally require larger samples. Binary-outcome calculations depend on assumed response proportions; time-to-event calculations depend especially on the required number of events, with recruitment and follow-up determining how many participants must enter. Allowance for attrition increases planned recruitment, but does not remove bias caused by missing outcomes. (fda.gov) For illustration, the normal approximation for two equally sized, independent groups with a continuous outcome is\n\n\\[n\\approx\\frac{2\\sigma^2(z_{1-\\alpha/2}+z_{1-\\beta})^2}{\\delta^2},\\]\n\nwhere \\(n\\) is the number per group, \\(\\sigma\\) the common standard deviation, \\(\\delta\\) the target mean difference, and \\(z_q\\) a standard-normal quantile. Under this approximation, detecting a difference of 5 units with \\(\\sigma=10\\), two-sided \\(\\alpha=0.05\\), and 80% power requires approximately 63 evaluable participants per group. Inflating this illustrative requirement for 15% attrition gives \\(\\lceil63/0.85\\rceil=75\\) recruits per group. (ncbi.nlm.nih.gov)\n\nMultiplicity arises when success can be declared through several outcomes, treatment comparisons, or repeated examinations of accumulating data. Without adjustment, these opportunities can inflate the probability of at least one false-positive conclusion. Prespecified procedures include the Bonferroni correction, which allocates significance thresholds across tests, and hierarchical testing, which permits later claims only after earlier hypotheses pass. Conversely, requiring all co-primary outcomes to succeed does not create the same inflation as allowing any one to establish success. The FDA’s October 2022 guidance addresses these endpoint structures; repeated interim testing additionally requires coordinated significance boundaries or alpha-spending procedures. (fda.gov)\n\nFrequentist methods evaluate uncertainty through the behavior of procedures under hypothetical repeated sampling, controlling type I and type II errors. Bayesian inference combines a prior distribution with the observed-data likelihood to obtain a posterior distribution. A Bayesian success rule might require a high posterior probability that the treatment benefit exceeds a clinically specified threshold. Prior information may come from earlier studies, but its relevance and influence require explicit justification and sensitivity analysis. Bayesian sample-size planning commonly uses simulation to examine decision probabilities under plausible scenarios, including misleading historical information; using posterior probabilities does not automatically guarantee acceptable frequentist error rates. The FDA’s 2010 medical-device guidance describes both prior specification and simulation-based assessment of operating characteristics. (fda.gov)"
    },
    {
      "heading": "Estimands, Analysis Populations, and Missing Data",
      "text": "An estimand specifies the treatment effect that a trial aims to quantify, rather than the statistical procedure used to estimate it. The International Council for Harmonisation’s ICH E9(R1) addendum, adopted in 2019, defines five attributes: the treatment conditions being compared; the target population; the outcome variable, including its assessment time; how intercurrent events are handled; and the population-level summary, such as a difference in means or risks. These attributes distinguish questions that otherwise appear identical. For example, a trial comparing two diabetes treatments could target the between-group difference in mean glycated haemoglobin at 26 weeks either regardless of rescue medication or under a hypothetical scenario in which rescue medication was unavailable. Those are different estimands, even with the same participants and scheduled measurements. (database.ich.org)\n\nIntercurrent events occur after treatment initiation and affect the interpretation or existence of measurements relevant to the research question. Examples include treatment discontinuation, rescue medication, switching to another treatment, and death. ICH E9(R1) describes five strategies: treatment policy uses outcomes regardless of the event; hypothetical targets outcomes under a specified scenario in which the event would not occur; composite incorporates the event into the outcome, for example classifying rescue treatment as failure; while-on-treatment considers outcomes before the event; and principal stratum targets participants defined by whether the event would occur under the treatments being compared. Different strategies may address different events within one estimand. (database.ich.org)\n\nAnalysis populations identify whose data enter an analysis, but their labels do not completely define its research question. The intention-to-treat principle includes all randomized participants and analyzes them in their assigned groups, regardless of adherence, preserving the comparison established by randomization. ICH E9’s “full analysis set” approximates that ideal, with any exclusions requiring justification. A per-protocol population restricts analysis to participants meeting prespecified requirements for treatment exposure, measurements, and absence of major protocol deviations. Such selection can introduce bias because adherence may depend on prognosis or treatment effects; it does not automatically estimate the causal effect of full adherence. Safety analyses commonly include participants receiving at least one dose and group them by treatment actually received, addressing exposure-related harms rather than the efficacy question addressed by randomized assignment. (database.ich.org)\n\nIntercurrent events are not themselves missing observations. A participant may discontinue medication yet continue all outcome assessments; conversely, someone continuing treatment may miss a visit. Under a treatment-policy strategy, post-discontinuation measurements remain relevant, so failure to collect them creates missing data. Death requires particular care: a symptom or quality-of-life score after death does not exist, rather than merely remaining unrecorded. A suitable estimand might incorporate death into a composite outcome or use another explicitly defined strategy. Retaining every participant in an analysis population does not, by itself, resolve these measurement problems. (database.ich.org)\n\nMissing-data methods depend on assumptions about why observations are absent. Missing completely at random means missingness is unrelated to observed or unobserved values; missing at random permits dependence on observed information but not on the missing values after conditioning on that information. Missing not at random allows remaining dependence on unobserved values—for example, unrecorded symptom scores being worse even after accounting for recorded history. Multiple imputation generates several plausible completed datasets, analyzes each, and combines results while incorporating uncertainty about the missing values. Its validity depends on the imputation model and assumptions; neither imputation nor a repeated-measures model automatically eliminates bias. (ema.europa.eu)\n\nPrespecified sensitivity analyses examine whether conclusions about the same estimand withstand plausible departures from the primary analysis assumptions. For example, missing outcomes can be shifted systematically toward worse values to determine when a conclusion changes—a tipping-point analysis. Analyses targeting a different estimand instead provide supplementary answers to different questions. The distinction prevents apparent agreement between methods from being mistaken for robustness when the methods actually estimate different treatment effects. (ema.europa.eu)"
    },
    {
      "heading": "Adaptive Designs and Master Protocols",
      "text": "An adaptive design permits prospectively planned modifications based on accumulating trial data. Rules specify when interim analyses occur, which data inform decisions, and which changes are permitted. Sample-size reassessment may use updated estimates of outcome variability or event rates; treatment-arm selection may discontinue less promising doses while retaining others. Prespecified boundaries can stop enrollment early for efficacy or futility—the latter indicating that success under the planned analysis is unlikely, rather than proving absence of benefit. These procedures differ from unplanned changes chosen after inspecting favorable results. The United States Food and Drug Administration’s December 2019 guidance describes their design and evaluation. (fda.gov)\n\nAdaptation requires statistical and operational safeguards. Simulations examine false-positive rates, power, sample sizes, and selection probabilities under different treatment effects and recruitment patterns. Repeated efficacy testing requires methods that control type I error, such as prespecified sequential boundaries. Access to unblinded interim comparisons is restricted, commonly to an independent monitoring committee and supporting statisticians, while investigators remain unaware of comparative results. Decision rules and communications are documented in advance. Effect estimates and confidence intervals may require adjustment for selection or early stopping; an ordinary fixed-sample analysis can otherwise misrepresent uncertainty. (fda.gov)\n\nA master protocol coordinates multiple substudies through a common framework, potentially sharing screening, sites, outcome assessments, and controls. Its main forms describe different organizational features. A basket trial evaluates a treatment across multiple diseases or disease subtypes, often linked in oncology by a molecular alteration rather than the tumor’s organ of origin. An umbrella trial evaluates several treatments within one disease, frequently assigning participants to substudies according to a biomarker. A platform trial evaluates treatments on an ongoing basis, allowing arms to enter or leave over time. These categories can overlap: a platform may contain basket or umbrella components. A master protocol is therefore not synonymous with an adaptive statistical design. (fda.gov)\n\nOncology illustrates these distinctions. NCI-MATCH, opened in August 2015, comprised molecularly defined substudies testing targeted treatments across tumor types. It ultimately enrolled 1,201 participants into 38 treatment arms; each molecular-treatment substudy represented a basket-style question rather than a comparison restricted to one organ-specific cancer. Lung-MAP, launched in 2014, used an umbrella framework initially focused on advanced squamous non-small cell lung cancer. Tumor testing directed participants toward relevant treatment substudies, with non-match studies available for some participants lacking an actionable match. Its framework allowed new treatment groups to be added and completed groups to close. These examples distinguish molecular screening and treatment assignment from randomization, which depends on the individual substudy’s design. (pmc.ncbi.nlm.nih.gov)\n\nShared controls can reduce duplication because the same eligible control participants contribute to several treatment comparisons. However, eligibility and calendar time determine which comparisons preserve the protection of randomization. For an arm introduced later, concurrent controls are recruited while that treatment is available; nonconcurrent controls include participants recruited outside that interval. Changes in prognosis, diagnostic practice, background care, or recruitment can confound comparisons with earlier controls. FDA’s draft master-protocol guidance therefore favors primary comparisons using concurrently randomized participants eligible for the treatment concerned. Borrowing earlier controls requires explicit assumptions, methods addressing temporal changes, and sensitivity analyses, including comparison with concurrent-only results. Adding an arm does not make the platform’s entire historical control population automatically comparable. (fda.gov)"
    },
    {
      "heading": "Ethics and Informed Consent",
      "text": "The 1979 Belmont Report identifies three principles of research ethics: respect for persons, beneficence, and justice. Respect for persons requires recognition of individual autonomy and protection for people with diminished autonomy; beneficence requires minimizing harm and assessing anticipated benefits against risks; justice concerns fair distribution of research burdens and benefits. Thus, recruiting disadvantaged groups merely because they are readily accessible can be unjust even when individuals consent. Scientific validity is also an ethical requirement: exposing participants to risk requires a design capable of producing valuable knowledge. The World Medical Association’s 2024 Declaration of Helsinki requires prior approval by an independent research ethics committee, which examines the protocol, participant protections, and risk–benefit justification. Consent does not replace this review. (hhs.gov)\n\nInformed consent is an ongoing communication process, not simply a signed form. Before enrollment, prospective participants receive understandable information about the research purpose, procedures, duration, foreseeable risks, possible benefits, alternatives, confidentiality, and relevant costs or payments. They need adequate time to consider participation and ask questions, without coercion or undue influence. Under United States FDA guidance, participation must be voluntary, and refusal or withdrawal must not entail penalties or loss of benefits to which the person is otherwise entitled. Information that could affect willingness to continue, such as newly identified risks, must be communicated during the study. Withdrawal from treatment can be distinguished from agreement to continued follow-up; previously collected data may remain available for analysis under applicable requirements. (hhs.gov)\n\nClinical equipoise, formulated by Benjamin Freedman in 1987, locates uncertainty about the preferred treatment within the expert clinical community rather than requiring every investigator to have no personal preference. It provides an ethical rationale for comparing treatments when their relative merits remain genuinely unsettled. Control-group selection nevertheless requires separate justification. The 2024 Declaration of Helsinki generally calls for comparison with the best proven intervention. Placebo or no intervention is acceptable when no proven intervention exists, or when compelling methodological reasons require it and withholding the best proven intervention creates no additional risk of serious or irreversible harm. (pubmed.ncbi.nlm.nih.gov)\n\nA distinct consent problem is therapeutic misconception: misunderstanding research as care organized solely around the individual participant’s interests. For example, a participant may believe that randomization assigns the treatment the physician considers best for that particular person, rather than allocating treatment according to a research design. Explaining the study’s knowledge-producing purpose, allocation procedures, and uncertainty about personal benefit addresses this distinction. Hope for improvement alone does not establish misconception; the central issue is whether the participant understands how research differs from individualized treatment. (pmc.ncbi.nlm.nih.gov)\n\nLimited consent capacity requires additional safeguards rather than automatic exclusion. The Declaration of Helsinki provides for permission from a legally authorized representative and involvement of the participant to the extent possible, including assent and respect for dissent. In United States research involving children, ethics committees assess capacity for assent using age, maturity, and psychological state; parental permission and the child’s affirmative agreement serve different functions. If an enrolled child reaches the legal age of consent, consent for continuing research interactions or interventions generally must be obtained unless an ethics committee approves a waiver. (wma.net)\n\nProtection also extends to privacy and confidentiality of personal information, and to arrangements for research-related injury. The Declaration of Helsinki requires appropriate treatment and compensation for participants harmed through research. Requirements differ across jurisdictions: for United States FDA-regulated research involving more than minimal risk, consent information must explain whether compensation and medical treatment are available for injury and where further details can be obtained. This disclosure requirement does not itself establish universal entitlement to compensation. (wma.net)"
    },
    {
      "heading": "Safety Assessment and Independent Monitoring",
      "text": "An adverse event is an unfavorable medical occurrence during investigational treatment, whether or not the intervention caused it; it may be a symptom, disease, or abnormal laboratory finding. An adverse reaction implies at least a reasonable possibility of a causal relationship. Seriousness concerns outcomes or medical consequences: death, immediate risk of death, inpatient hospitalization or its prolongation, persistent or significant incapacity, congenital anomaly, or another medically important event requiring intervention to prevent such consequences. Severity describes intensity. A severe headache need not meet seriousness criteria, whereas an event with relatively mild symptoms may be serious because it requires hospitalization. “Life-threatening” means that death was a risk when the event occurred, not that a hypothetical worse episode could have been fatal. (fda.gov)\n\nCausality assessment considers the timing of exposure, alternative explanations such as underlying illness and concomitant medicines, and changes after treatment withdrawal or re-exposure. Temporal association alone does not establish causation. Investigators document their assessment, while sponsors review individual cases alongside accumulating evidence. Some risks become apparent only through an excess of events in one treatment group rather than through an unmistakable individual reaction. Expectedness is assessed separately against applicable reference safety information: a known reaction occurring with greater specificity or severity may qualify as unexpected. These distinctions determine which reports require expedited communication, rather than treating every serious event as a drug reaction. (gov.uk)\n\nProtocols specify safety assessments and reporting periods, including symptom collection, examinations, and laboratory testing. Records preserve the reported event, onset and resolution, intensity, seriousness, treatment action, outcome, and causal assessment. Standard terminology such as the Medical Dictionary for Regulatory Activities (MedDRA) supports consistent coding and grouped review without replacing the original clinical description. Investigators generally report serious adverse events immediately to the sponsor, followed by additional information; prespecified exceptions may apply to events collected as trial endpoints. Sponsors evaluate the cumulative safety profile and communicate important new risks to investigators, ethics committees, and regulators as applicable. Such findings can require changes to monitoring procedures, the protocol, or informed-consent information. (gov.uk)\n\nAn independent data monitoring committee (DMC), also called a data and safety monitoring board, periodically examines accumulating safety and efficacy evidence and recommends continuation, modification, or stopping. It does not replace investigators’ immediate clinical responsibilities. Clinical and statistical expertise are central to its membership, with independence supported by conflict-of-interest assessment. A charter agreed before enrollment defines responsibilities, meeting schedules, access to data, interim monitoring procedures, and communication pathways. Confidential review of unmasked treatment-group results allows the committee to compare benefits and harms while limiting disclosure to personnel whose conduct of the trial could be influenced by interim findings. Recommendations ordinarily go to sponsor leadership or a steering committee authorized to act on them. (ctti-clinicaltrials.org)\n\nBenefit–risk monitoring is not restricted to detecting toxicity. The Systolic Blood Pressure Intervention Trial (SPRINT), involving 9,361 participants, illustrates a decision driven by efficacy despite increased rates of some adverse events with intensive treatment. Its independent monitoring board reviewed interim outcomes and adverse events. The blood-pressure intervention stopped in August 2015, after a median follow-up of 3.26 years, when cardiovascular outcomes crossed the efficacy boundary at two consecutive assessments. The decision concerned the balance of observed benefit and harm, not a finding that intensive treatment was free of risk. (nhlbi.nih.gov)\n\nStopping treatment for one participant is distinct from suspending or terminating a trial. Individual treatment discontinuation may leave consented safety and outcome follow-up in place. A temporary suspension pauses specified activities, such as recruitment or dosing, while a concern is investigated; termination permanently ends the trial prematurely. Neither removes obligations toward enrolled participants. ICH E6(R3), issued as FDA guidance in September 2025, calls for prompt notification of participants, appropriate therapy and follow-up, and notification of investigators, regulators, and ethics committees when a trial is suspended or prematurely terminated. (fda.gov)"
    },
    {
      "heading": "Trial Organization, Funding, and Regulation",
      "text": "The sponsor takes responsibility for initiating and managing a trial and arranging its financing; it need not supply all the money itself. Sponsors may be companies, universities, public agencies, or other organizations. Their functions include selecting qualified investigators, establishing operational arrangements, providing investigational products, overseeing service providers, and meeting applicable reporting obligations. A contract research organization (CRO) performs contracted activities such as site management, monitoring, or statistical analysis. Written agreements specify transferred activities, but outsourcing does not remove the sponsor’s overarching responsibility for participant protection and reliable results. An investigator who also initiates and sponsors a study assumes both sets of responsibilities. (fda.gov)\n\nThe investigator, usually led at each site by a principal investigator, is responsible for protocol-compliant conduct, supervision of delegated activities, participant medical care, and control of investigational drugs. Research sites provide the personnel and facilities for recruitment, intervention delivery, examinations, and follow-up; investigators must ensure appropriate informed consent and required ethics-review arrangements. In multicenter research, a coordinating center supplies shared operational or analytical infrastructure rather than replacing local investigator responsibility. Its assigned functions may include communication among sites, centralized reporting, and administration of study-wide policies—for example, the US National Institute of Diabetes and Digestive and Kidney Diseases directs participating investigators to submit study financial-disclosure forms to the coordinating center. (fda.gov)\n\nFunding can combine commercial support, public grants, charitable contributions, and institutional infrastructure. In a collaborative example, the RECOVERY trial’s March 15, 2022 protocol identified grants to the University of Oxford from UK Research and Innovation, the National Institute for Health Research, and Wellcome, alongside core support from several public and philanthropic organizations. Financial support and regulatory sponsorship are therefore distinct roles. Financial relationships also require disclosure and, where necessary, management: under US National Institutes of Health rules, institutions review investigators’ significant financial interests, determine whether they could directly and significantly affect research design, conduct, or reporting, and manage and report identified conflicts of interest. Separately, FDA regulations require financial information about investigators whose covered studies support marketing applications; relevant changes must be reported during the study and for one year after completion. Disclosure alone is not equivalent to a determination that a conflict has been adequately managed. (recoverytrial.net)\n\nGood Clinical Practice (GCP) provides an ethical, scientific, and quality framework for conducting trials and producing reliable results. The International Council for Harmonisation’s E6(R3) principles and Annex 1 reached Step 4 adoption on January 6, 2025; FDA issued its final E6(R3) guidance in September 2025. The revision emphasizes proportionate oversight, clear allocation of responsibilities, and quality built into trial design. International harmonization does not create a single worldwide clinical-trial law: implementation occurs through regional and national systems. FDA guidance describes the agency’s current thinking and is generally nonbinding, whereas applicable statutes and regulations impose legal duties. In the United States, relevant requirements include 21 CFR Parts 50 and 56 for participant protection and institutional review boards, Part 54 for financial disclosure, and Part 312 for investigational drugs. (fda.gov)\n\nUnited States example—investigational new drugs. Where an investigational new drug (IND) application is required, the sponsor submits animal pharmacology and toxicology findings, manufacturing information, proposed clinical protocols, and investigator information to FDA. An initial IND normally becomes effective 30 days after FDA receives it unless the agency imposes a clinical hold; FDA may also notify the sponsor earlier that the investigation may proceed. Required institutional review board approval remains a separate condition. An effective IND permits investigational use under the applicable framework, not commercial marketing: it supplies an exemption allowing investigational distribution, rather than establishing that the product has demonstrated safety and effectiveness for sale. Marketing requires a separate applicable approval or licensure process. (fda.gov)"
    },
    {
      "heading": "Data Management and Quality Assurance",
      "text": "Source records preserve the original observations underlying trial results: medical notes, laboratory reports, imaging files, participant diaries, and electronically recorded measurements. A case-report form (CRF) organizes the protocol-required information reported to the sponsor; it need not reproduce the entire medical record. Electronic data capture systems receive information through manual entry, direct recording, or transfers from other systems. When an observation is entered directly into an electronic CRF, that record may itself constitute the source. The FDA’s September 2013 guidance on electronic source data emphasizes identifying authorized data originators and maintaining traceability between source observations and submitted data. (fda.gov)\n\nData cleaning combines automated checks with clinical review. Checks can flag missing fields, inconsistent dates, or implausible values; a data query requests clarification rather than automatically establishing that an observation is wrong. Coding converts reported terms into standardized categories for analysis while retaining their original wording. Before database finalization, responsible personnel document query resolution, coding completion, reconciliation of serious adverse-event information, and relevant quality-control checks. Finalized datasets are protected against uncontrolled editing; subsequent changes require documented authorization and traceability. (fda.gov)\n\nTraceability allows an analyst or reviewer to reconstruct how an observation became an analysis value. An audit trail records the creation, modification, or deletion of electronic records, including the responsible user, date and time, old and new values, and, where applicable, the reason for change. Corrections must not obscure previously recorded information. The FDA’s October 2024 electronic-records guidance recommends protecting audit trails against alteration or disabling. Access controls assign permissions according to responsibilities, using individual accounts rather than shared credentials. Encryption, secure backups, recovery procedures, and retention of relevant metadata support data privacy and record integrity. These safeguards must also cover externally hosted systems; outsourcing storage does not remove the regulated organization’s responsibilities. (fda.gov)\n\nMonitoring, auditing, and inspection serve different functions. Monitoring is continuing quality control, combining site review, remote access, and centralized analysis to identify problems during conduct. A quality-assurance audit independently evaluates trial processes and compliance, separately from routine monitoring. A regulatory inspection is an official examination by a competent authority. Centralized monitoring can identify site-specific or systemic anomalies and direct targeted follow-up; it is not synonymous with checking every recorded value against its source. (fda.gov)\n\nQuality by design moves attention upstream, before errors accumulate. ICH E8(R1), finalized in 2021, calls for identifying “critical to quality” factors during planning: attributes whose failure would compromise participant protection or the reliability and interpretation of results. Examples include workable consent procedures, reliable primary-outcome measurement, and operational arrangements that minimize avoidable loss of essential data. Risk-proportionate oversight allocates effort according to the likelihood and consequences of failure rather than treating every field as equally important. Unnecessary measurements and complicated workflows can themselves create errors and burdens. The approach therefore includes simplifying forms, testing feasibility, training staff on critical procedures, and revisiting controls as experience reveals new risks. (database.ich.org)"
    },
    {
      "heading": "Registration, Reporting, and Data Sharing",
      "text": "Prospective trial registration creates a public record before recruitment of the first participant. The record identifies the sponsor, interventions, eligibility criteria, planned enrollment, outcomes, and measurement time points, providing a dated account against which subsequent reports can be compared. Registries assign unique identifiers that link records to publications and other documents. The World Health Organization International Clinical Trials Registry Platform (ICTRP) is not itself a registry: it establishes international standards and provides a search portal aggregating records from participating data providers. Because one trial may appear in several registries, cross-referencing identifiers helps distinguish multiple records of the same study from separate studies. (who.int)\n\nRegistration and results disclosure are distinct activities. A registration entry describes the planned study and its status; posted summary results provide structured aggregate findings, including participant flow, baseline characteristics, outcome data, and adverse events. A journal article presents methods, findings, and interpretation, usually following peer review. A clinical study report is a more extensive technical account used in regulatory submissions, with supporting materials such as the protocol and statistical documentation. Participant-facing summaries explain the study and its overall findings in nontechnical language, including adverse reactions and their frequency. These formats serve different audiences and do not substitute for one another. The WHO-led 2017 joint statement endorsed disclosure of summary results within 12 months of primary study completion as a global norm, with journal publication within an indicative 24 months; these are policy expectations rather than a single worldwide legal deadline. (grants.nih.gov)\n\nCONSORT 2025, published on April 14, 2025, replaced CONSORT 2010 with a 30-item checklist and participant-flow diagram for reporting randomized trials. The diagram documents enrollment, allocation, follow-up, and analysis, making losses and exclusions visible by treatment group. The checklist calls for reporting how harms were assessed and analyzed, as well as harms observed in each group. It also addresses how interventions were actually administered, important changes after trial commencement, and outcomes or analyses that were not prespecified, with reasons for those changes. Its open-science items identify the registration record, access to the protocol and statistical analysis plan, and arrangements for sharing data and code. CONSORT specifies minimum reporting information; checklist completion is not certification that a trial’s design or conduct was valid. (bmj.com)\n\nPublication bias arises when the availability of whole studies depends on their findings; selective outcome reporting occurs when the outcomes or analyses presented are chosen according to their results. In a 2008 study, Erick H. Turner and colleagues compared regulatory reviews with publications for 74 antidepressant trials involving 12,564 participants. Thirty-one percent of the trials were unpublished; the journal literature presented 94% as positive, compared with 51% classified as positive in the regulatory assessment. Public registration makes otherwise unpublished studies identifiable, while access to dated protocols, amendments, and analysis plans allows readers to examine outcome switching and distinguish prespecified from exploratory analyses. Neither registration alone nor publication alone ensures complete disclosure. (nejm.org)\n\nParticipant-level data sharing provides access beyond aggregate tables but need not mean unrestricted public release. Under International Committee of Medical Journal Editors requirements, trial manuscripts submitted from July 1, 2018 must include a data-sharing statement, and trials beginning enrollment from January 1, 2019 must register a sharing plan. These requirements mandate disclosure of arrangements, not sharing itself. Statements specify which deidentified data and supporting documents are available, when, to whom, and under what conditions. Controlled-access systems combine scientific review with privacy safeguards: the Yale Open Data Access Project, for example, reviews research proposals, requires a data-use agreement, and provides approved access through a secure platform or secure transfer, depending on the data partner. Supporting protocols and analysis documentation accompany data to enable interpretable secondary analyses. (pmc.ncbi.nlm.nih.gov)"
    },
    {
      "heading": "Interpretation and Use of Trial Evidence",
      "text": "Statistical significance does not establish clinical importance. A P value describes the compatibility of the observations with a specified statistical model and null hypothesis; it does not measure the size of a benefit or the probability that treatment works. Interpretation therefore emphasizes the effect estimate and its confidence interval, considered against a clinically meaningful difference. A large trial may detect a trivial improvement, whereas a smaller trial may yield an interval encompassing substantial benefit, no effect, and harm. Failure to reject the null hypothesis is not proof of equivalence. Confidence intervals quantify sampling uncertainty under the analysis assumptions, not every possible source of bias. (cochrane.org)\n\nAbsolute and relative effects answer different questions. In the 2015 EMPA-REG OUTCOME trial, the primary cardiovascular composite outcome occurred in 490 of 4,687 participants receiving empagliflozin (10.5%) and 282 of 2,333 receiving placebo (12.1%). The reported hazard ratio was 0.86, with a 95.02% confidence interval of 0.74–0.99. The observed proportions differed by approximately 1.6 percentage points; the hazard ratio instead summarized the relative event hazard over follow-up and was not a percentage-point reduction. (nejm.org) More generally, an assumed constant risk ratio of 0.80 would reduce a baseline risk of 20% to 16%, but a baseline risk of 2% to 1.6%: absolute reductions of 4 and 0.4 percentage points, respectively. Applying relative effects to another population therefore requires an appropriate baseline risk and time horizon. (cochrane.org)\n\nInternal validity concerns whether the estimated comparison credibly reflects the intervention’s effect within the study. Randomization supports that inference, but interpretation also examines departures from assigned treatment, missing outcomes, measurement procedures, and selective analyses. Prespecified primary analyses and sensitivity analyses help establish whether conclusions depend on particular assumptions. Generalizability is a separate question: the relevance of the result to patients, treatment settings, doses, and follow-up periods outside the trial. Exclusion of patients with substantial comorbidity, unusually intensive clinical supervision, or a comparator unlike current practice can limit applicability without necessarily invalidating the trial’s internal comparison. (fda.gov)\n\nSubgroup claims require evidence that treatment effects differ, not merely that one subgroup has a small P value and another does not. A direct test of interaction compares the effects between groups. Credibility increases when the characteristic and analysis were prespecified, the proposed difference has a plausible rationale, and independent evidence supports it. Searching many age categories, biomarkers, or geographic groups increases opportunities for chance findings. Treatment-effect heterogeneity also depends on the measurement scale: differing absolute benefits can arise from differing baseline risks even when relative effects are similar. An overall average does not establish identical effects for every participant, but exploratory subgroup estimates do not by themselves establish individualized treatment rules. (cochrane.org)\n\nIndividual trials enter systematic reviews through explicit eligibility criteria, systematic identification, and assessment of bias. Where studies address sufficiently comparable questions, meta-analysis combines effect estimates while accounting for their precision; differences among populations, interventions, and methods remain relevant to interpretation. The GRADE framework assesses certainty separately for important outcomes, considering risk of bias, inconsistency, indirectness, imprecision, and publication bias. Clinical guidelines use this evidence alongside benefits, harms, patient preferences, and resource considerations. Product assessments likewise examine the strength of the evidence rather than treating a statistically significant result as automatically sufficient; the FDA’s September 2023 draft guidance described circumstances in which one adequate, well-controlled investigation plus confirmatory evidence could establish substantial evidence of effectiveness. (cochrane.org)\n\nUncommon or delayed harms may remain unresolved even after convincing efficacy findings. Trials with few exposed participants generate little information about rare events, while short follow-up cannot characterize effects emerging years later. Evidence synthesis may therefore incorporate longer-term follow-up and nonrandomized studies for harms inadequately captured in randomized trials, with attention to confounding and differences between populations. A single positive trial can leave uncertainty about reproducibility, effect magnitude, applicability, and long-term safety; additional evidence may strengthen, narrow, or alter the original interpretation. (cochrane.org)"
    },
    {
      "heading": "Decentralized Trials and Sustainable Research Infrastructure",
      "text": "Decentralized elements move some trial activities from specialist research sites to participants’ homes, local clinics, or other convenient locations. They include video consultations, blood collection by visiting nurses, testing at local laboratories, electronic symptom diaries, and wearable sensors. Decentralization need not be complete: a hybrid trial may retain hospital-based treatment administration while conducting follow-up remotely. The United States Food and Drug Administration’s September 2024 guidance explicitly accommodates this combination. Suitability depends on the intervention and procedures; complex assessments or treatments requiring immediate specialist support may remain site-based. Decentralization changes where activities occur, not the scientific requirements for a valid comparison or the investigator’s responsibility for participants. (fda.gov)\n\nRemote participation can reduce travel but can also reproduce the digital divide. Smartphone ownership, connectivity, language, digital literacy, and accessibility affect who can participate and whether measurements are completed. FDA’s December 2023 guidance on digital health technologies addresses provision of suitable equipment and telecommunications access, participant training, and technical support. Measurement consistency requires more than distributing identical devices: sensors and software must be fit for the intended population and use, with appropriate verification, validation, and usability evaluation. Protocols must address device failures, software changes, and missing transmissions. A movement sensor, for example, must measure the intended activity reliably under the conditions in which participants actually use it, rather than merely perform accurately in a laboratory. (fda.gov)\n\nDispersed procedures also require explicit operational coordination. Investigators need access to information collected by local providers, clear allocation of responsibilities, and procedures for escalating abnormal findings or adverse events. FDA’s decentralized-trial guidance distinguishes routine services performed by local healthcare professionals from trial-specific activities requiring additional training. Procedures vulnerable to operator variability, such as spirometry, may require training or video supervision. Safety plans must identify how participants contact the research team and obtain urgent local care; remote data collection does not itself provide continuous clinical surveillance. (fda.gov)\n\nRoutine health records can support recruitment, baseline characterization, and outcome follow-up, but their evidential role depends on study design. A randomized controlled trial remains randomized when hospital admissions or deaths are obtained through record linkage. Conversely, comparing treatments selected during routine care remains vulnerable to confounding, even in a very large database. FDA guidance on electronic health records and claims data emphasizes whether exposures, outcomes, and relevant covariates are captured adequately. Coding differences, care received outside the contributing system, and incompletely recorded deaths can distort results. Thus, electronic health records and real-world data are potential data sources, not substitutes for a defensible treatment-comparison design. (fda.gov)\n\nThe RECOVERY trial illustrates integration with an existing health system during an emergency, rather than a wholly remote trial. Established in March 2020 across United Kingdom National Health Service hospitals, it had enrolled more than 11,500 participants at over 175 hospitals by June 16, 2020. Its initial dexamethasone comparison included 2,104 participants assigned treatment and 4,321 assigned usual care. The investigators’ methodological account describes short web-based forms supplemented by centrally held hospital, intensive-care, and death-registration records; primary-outcome follow-up exceeded 99%. Existing clinical services and linked information systems supported rapid recruitment without abandoning random allocation. (recoverytrial.net)\n\nDurable infrastructure extends beyond software to trained research teams, coordinating centres, laboratories, community relationships, and adequately resourced ethics and regulatory bodies. The World Health Organization’s September 2024 guidance advocates sustained domestic financing, research career development, and integration of trials into national health systems. It also supports regional and international collaboration, coordinated review processes, and proportionate oversight. Its preparedness model keeps research networks functional through studies addressing endemic conditions, allowing them to pivot during epidemics rather than rebuilding staffing, partnerships, and operational systems after an emergency begins. (who.int)"
    }
  ],
  "links": [
    "vaccine",
    "public-health",
    "experimental-design",
    "randomized-controlled-trial",
    "informed-consent",
    "world-health-organization",
    "randomization",
    "epidemiology",
    "surgery",
    "quality-of-life",
    "cancer",
    "tuberculosis",
    "ethics",
    "pharmacology",
    "pharmacokinetics",
    "biomarker",
    "placebo",
    "diagnosis",
    "disability",
    "pregnancy",
    "confounding",
    "factorial-design",
    "statistical-interaction",
    "confidence-interval",
    "effect-size",
    "significance-level",
    "statistical-power",
    "multiple-testing",
    "bonferroni-correction",
    "type-i-and-type-ii-errors",
    "bayesian-inference",
    "prior-distribution",
    "posterior-distribution",
    "sensitivity-analysis",
    "estimator-bias",
    "missing-data-imputation",
    "data-privacy",
    "causation",
    "database",
    "peer-review",
    "p-value",
    "null-hypothesis",
    "digital-divide",
    "health-system"
  ],
  "sources": [
    {
      "title": "Clinical trials",
      "url": "https://www.who.int/news-room/questions-and-answers/item/clinical-trials"
    },
    {
      "title": "NIH's Definition of a Clinical Trial",
      "url": "https://grants.nih.gov/policy-and-compliance/policy-topics/clinical-trials/definition"
    },
    {
      "title": "Observational study - Toolkit",
      "url": "https://toolkit.ncats.nih.gov/glossary/observational-study/"
    },
    {
      "title": "What Are Clinical Trials? - NCI",
      "url": "https://www.cancer.gov/research/participate/clinical-trials/what-are-clinical-trials"
    },
    {
      "title": "Clinical Research and Cancer - NCI",
      "url": "https://www.cancer.gov/research/participate/about-cancer-clinical-research"
    },
    {
      "title": "Clinical Imaging Trials Basics - NCI",
      "url": "https://dctd.cancer.gov/research/research-areas/imaging/basics/trials"
    },
    {
      "title": "Primary purpose - Toolkit",
      "url": "https://toolkit.ncats.nih.gov/glossary/primary-purpose/"
    },
    {
      "title": "James Lind and scurvy: 1747 to 1795.",
      "url": "https://www.jameslindlibrary.org/articles/james-lind-and-scurvy-1747-to-1795/"
    },
    {
      "title": "Lind J (1753)",
      "url": "https://www.jameslindlibrary.org/lind-j-1753/"
    },
    {
      "title": "Who was James Lind, and what exactly did he achieve?",
      "url": "https://www.jameslindlibrary.org/articles/who-was-james-lind-and-what-exactly-did-he-achieve/"
    },
    {
      "title": "Medical Research Council (1948)",
      "url": "https://www.jameslindlibrary.org/medical-research-council-1948b/"
    },
    {
      "title": "The MRC randomized trial of streptomycin and its legacy: a view from the clinical front line.",
      "url": "https://www.jameslindlibrary.org/articles/the-mrc-randomized-trial-of-streptomycin-and-its-legacy-a-view-from-the-clinical-front-line/"
    },
    {
      "title": "45 CFR 46 FAQs",
      "url": "https://www.hhs.gov/ohrp/regulations-and-policy/guidance/faq/45-cfr-46/index.html"
    },
    {
      "title": "Commemorating the 50th Anniversary of the National Research Act",
      "url": "https://www.hhs.gov/sites/default/files/2024-oew-program-book-final.pdf"
    },
    {
      "title": "Read the Belmont Report",
      "url": "https://www.hhs.gov/ohrp/regulations-and-policy/belmont-report/read-the-belmont-report/index.html"
    },
    {
      "title": "Federal Policy for the Protection of Human Subjects ('Common Rule')",
      "url": "https://www.hhs.gov/ohrp/regulations-and-policy/regulations/common-rule/index.html"
    },
    {
      "title": "Exploratory IND Studies",
      "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/exploratory-ind-studies"
    },
    {
      "title": "Phases of Clinical Trials - RaDaR",
      "url": "https://registries.ncats.nih.gov/glossary/phases-of-clinical-trials/"
    },
    {
      "title": "ICH E8(R1): General Considerations for Clinical Studies",
      "url": "https://database.ich.org/sites/default/files/E8-R1_Guideline_Step4_2022_0204%20%281%29.pdf"
    },
    {
      "title": "Step 3: Clinical Research",
      "url": "https://www.fda.gov/patients/drug-development-process/step-3-clinical-research?trk=article-ssr-frontend-pulse_little-text-block"
    },
    {
      "title": "Drug Development and Review Definitions",
      "url": "https://www.fda.gov/drugs/investigational-new-drug-application-ind/drug-development-and-review-definitions"
    },
    {
      "title": "Protocol Registration Data Element Definitions for Interventional and Observational Studies",
      "url": "https://clinicaltrials.gov/policy/protocol-definitions?trk=article-ssr-frontend-pulse_little-text-block"
    },
    {
      "title": "FDA-NIH Terminology For Clinical Research",
      "url": "https://www.fda.gov/media/178477/download?attachment="
    },
    {
      "title": "Step 3: Clinical Research",
      "url": "https://www.fda.gov/patients/drug-development-process/step-3-clinical-research"
    },
    {
      "title": "E 9 Statistical Principles for Clinical Trials",
      "url": "https://www.ema.europa.eu/en/documents/scientific-guideline/ich-e-9-statistical-principles-clinical-trials-step-5_en.pdf"
    },
    {
      "title": "Statistical Principles for Clinical Trials",
      "url": "https://database.ich.org/sites/default/files/E9_Guideline.pdf"
    },
    {
      "title": "ICH E8(R1) Draft Guideline: General Considerations for Clinical Studies",
      "url": "https://admin.ich.org/sites/default/files/inline-files/E8-R1EWG_Step2_DraftGuideline_2019_0508.pdf"
    },
    {
      "title": "Eligibility Criteria",
      "url": "https://dtc.cancer.gov/patients/eligibility_criteria.htm"
    },
    {
      "title": "Trial Summary - Learn - NLST - The Cancer Data Access System",
      "url": "https://cdas.cancer.gov/learn/nlst/trial-summary/"
    },
    {
      "title": "Enhancing Participation in Clinical Trials — Eligibility Criteria, Enrollment Practices, and Trial Designs",
      "url": "https://www.fda.gov/media/190162/download"
    },
    {
      "title": "Guidance for best practices for clinical trials",
      "url": "https://iris.who.int/bitstream/handle/10665/378782/9789240097711-eng.pdf?sequence=1"
    },
    {
      "title": "Data Retention When Subjects Withdraw from FDA-Regulated Clinical Trials; Guidance for Sponsors, Clinical Investigators, and IRBs",
      "url": "https://www.fda.gov/media/75138/download"
    },
    {
      "title": "Inclusion Across the Lifespan in Human Subjects Research",
      "url": "https://grants.nih.gov/policy-and-compliance/policy-topics/inclusion/lifespan"
    },
    {
      "title": "Pregnant Women: Scientific and Ethical Considerations for Inclusion in Clinical Trials",
      "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/pregnant-women-scientific-and-ethical-considerations-inclusion-clinical-trials"
    },
    {
      "title": "New global guidance puts forward recommendations for more effective and equitable clinical trials",
      "url": "https://www.who.int/news/item/25-09-2024-new-global-guidance-puts-forward-recommendations-for-more-effective-and-equitable-clinical-trials"
    },
    {
      "title": "Implementing Criteria to Expand Clinical Trial Eligibility",
      "url": "https://dctd.cancer.gov/research/research-areas/clinical-trials/modernization/criteria-expand-trial-eligibility"
    },
    {
      "title": "E10 Choice of Control Group and Related Issues in Clinical Trials",
      "url": "https://www.fda.gov/media/71349/download"
    },
    {
      "title": "E9 Statistical Principles for Clinical Trials",
      "url": "https://www.fda.gov/media/71336/download"
    },
    {
      "title": "Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products",
      "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-design-and-conduct-externally-controlled-trials-drug-and-biological-products"
    },
    {
      "title": "Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products",
      "url": "https://www.fda.gov/media/164960/download"
    },
    {
      "title": "CONSORT 2025 explanation and elaboration: updated guideline for reporting randomised trials",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11995452"
    },
    {
      "title": "SPIRIT 2025 explanation and elaboration: Item 24a, Blinding status",
      "url": "https://www.consort-spirit.org/spiriteande24a/item-3c-role-of-sponsor-and-funders"
    },
    {
      "title": "Regulatory Fitness in Rare Disease Clinical Trials Virtual Workshop",
      "url": "https://www.fda.gov/media/160996/download"
    },
    {
      "title": "Clinical Study Designs for Surgical Ablation Devices for Treatment of Atrial Fibrillation - Guidance for Industry and Food and Drug Administration Staff",
      "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-study-designs-surgical-ablation-devices-treatment-atrial-fibrillation-guidance-industry-and"
    },
    {
      "title": "Patient-Reported Outcome Measures: Use in Medical Product Development to Support Labeling Claims",
      "url": "https://www.fda.gov/media/77832/download"
    },
    {
      "title": "Clinical Outcome Assessment (COA): Frequently Asked Questions",
      "url": "https://www.fda.gov/about-fda/clinical-outcome-assessment-coa-frequently-asked-questions"
    },
    {
      "title": "Surrogate Endpoint Resources for Drug and Biologic Development",
      "url": "https://www.fda.gov/drugs/development-resources/surrogate-endpoint-resources-drug-and-biologic-development"
    },
    {
      "title": "Validated Surrogate Endpoint — BEST (Biomarkers, EndpointS, and other Tools) Resource",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK453484/"
    },
    {
      "title": "Multiple Endpoints in Clinical Trials — Guidance for Industry",
      "url": "https://www.fda.gov/media/162416/download"
    },
    {
      "title": "FDA Clinical Review",
      "url": "https://www.accessdata.fda.gov/drugsatfda_docs/nda/2021/206073Orig1s017.pdf"
    },
    {
      "title": "Non-Inferiority Clinical Trials to Establish Effectiveness",
      "url": "https://www.fda.gov/media/78504/download"
    },
    {
      "title": "Table 5.4, Summary of formulae for calculating trial size requirements for comparison of two groups of equal size",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK305517/table/chapter-5-tableGroup-4/"
    },
    {
      "title": "Guidance for the Use of Bayesian Statistics in Medical Device Clinical Trials",
      "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/guidance-use-bayesian-statistics-medical-device-clinical-trials"
    },
    {
      "title": "ICH E9(R1) Guideline",
      "url": "https://database.ich.org/sites/default/files/E9-R1_Step4_Guideline_2019_1203.pdf"
    },
    {
      "title": "ICH E9(R1) Training Material",
      "url": "https://database.ich.org/sites/default/files/E9%28R1%29%20Training%20Material%20-%20PDF_0.pdf"
    },
    {
      "title": "Guideline on Missing Data in Confirmatory Clinical Trials",
      "url": "https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-missing-data-confirmatory-clinical-trials_en.pdf"
    },
    {
      "title": "Adaptive Designs for Clinical Trials of Drugs and Biologics",
      "url": "https://www.fda.gov/media/78495/download"
    },
    {
      "title": "Master Protocols for Drug and Biological Product Development",
      "url": "https://www.fda.gov/media/174976/download?attachment="
    },
    {
      "title": "Guidance Recap Podcast: Master Protocols for Drug and Biological Product Development",
      "url": "https://www.fda.gov/drugs/guidances-drugs/guidance-recap-podcast-master-protocols-drug-and-biological-product-development"
    },
    {
      "title": "National Cancer Institute Basket/Umbrella Clinical Trials: MATCH, LungMAP, and Beyond",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6658131/"
    },
    {
      "title": "NCI-MATCH Precision Medicine Clinical Trial",
      "url": "https://www.cancer.gov/research/infrastructure/clinical-trials/featured/nci-match"
    },
    {
      "title": "Lung-MAP Clinical Trial",
      "url": "https://www.cancer.gov/types/lung/research/lung-map"
    },
    {
      "title": "WMA Declaration of Helsinki – Ethical Principles for Medical Research Involving Human Participants",
      "url": "https://www.wma.net/policies-post/wma-declaration-of-helsinki/"
    },
    {
      "title": "Informed Consent Guidance for IRBs, Clinical Investigators, and Sponsors",
      "url": "https://www.hhs.gov/guidance/sites/default/files/hhs-guidance-documents/FDA/Informed-Consent-Guidance-for-Institutional-Review-Boards-Clinical-Investigators-and-Sponsors.pdf"
    },
    {
      "title": "Informed Consent FAQs",
      "url": "https://www.hhs.gov/ohrp/regulations-and-policy/guidance/faq/informed-consent/index.html"
    },
    {
      "title": "Withdrawal of Subjects from Research Guidance (2010)",
      "url": "https://www.hhs.gov/ohrp/regulations-and-policy/guidance/guidance-on-withdrawal-of-subject/index.html"
    },
    {
      "title": "Equipoise and the ethics of clinical research",
      "url": "https://pubmed.ncbi.nlm.nih.gov/3600702/"
    },
    {
      "title": "Declaration – WMA",
      "url": "https://www.wma.net/policy-types/declaration/"
    },
    {
      "title": "Clinical Trials and Medical Care: Defining the Therapeutic Misconception",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC2082641/"
    },
    {
      "title": "Key Information and Facilitating Understanding in Informed Consent",
      "url": "https://www.fda.gov/media/176663/download?trk=public_post_comment-text"
    },
    {
      "title": "Research with Children FAQs",
      "url": "https://www.hhs.gov/ohrp/regulations-and-policy/guidance/faq/children-research/index.html"
    },
    {
      "title": "Institutional Review Boards Frequently Asked Questions",
      "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/institutional-review-boards-frequently-asked-questions"
    },
    {
      "title": "E2A Clinical Safety Data Management: Definitions and Standards for Expedited Reporting",
      "url": "https://www.fda.gov/media/71188/download"
    },
    {
      "title": "Clinical trials for medicines: collection, verification and reporting of safety events",
      "url": "https://www.gov.uk/guidance/clinical-trials-for-medicines-collection-verification-and-reporting-of-safety-events"
    },
    {
      "title": "IND Application Reporting: IND Safety Reports",
      "url": "https://www.fda.gov/drugs/investigational-new-drug-application-ind/ind-application-reporting-ind-safety-reports"
    },
    {
      "title": "E6(R3) Good Clinical Practice Guidance for Industry | September 2025",
      "url": "https://www.fda.gov/media/169090/download"
    },
    {
      "title": "CTTI Recommendations: Data Monitoring Committees",
      "url": "https://ctti-clinicaltrials.org/about/ctti-projects/data-monitoring-committees/ctti-recommendations-data-monitoring-committees/"
    },
    {
      "title": "Guidance for Clinical Trial Sponsors: Establishment and Operation of Clinical Trial Data Monitoring Committees",
      "url": "https://www.fda.gov/media/75398/download?locale=us"
    },
    {
      "title": "Systolic Blood Pressure Intervention Trial (SPRINT) Study",
      "url": "https://www.nhlbi.nih.gov/science/systolic-blood-pressure-intervention-trial-sprint-study"
    },
    {
      "title": "BioLINCC: Systolic Blood Pressure Intervention Trial (SPRINT)",
      "url": "https://biolincc.nhlbi.nih.gov/studies/sprint/"
    },
    {
      "title": "IND Applications for Clinical Investigations: Overview",
      "url": "https://www.fda.gov/drugs/investigational-new-drug-application-ind/ind-applications-clinical-investigations-overview"
    },
    {
      "title": "IND Application Procedures: Investigator's Responsibilities",
      "url": "https://www.fda.gov/drugs/investigational-new-drug-application-ind/ind-application-procedures-investigators-responsibilities"
    },
    {
      "title": "Financial Disclosure for Multicenter Clinical Trials",
      "url": "https://www.niddk.nih.gov/research-funding/human-subjects-research/policies-clinical-researchers/financial-disclosure-multicenter-clinical-trials"
    },
    {
      "title": "RECOVERY Protocol, Version 23.1, March 15, 2022",
      "url": "https://www.recoverytrial.net/files/recovery-protocol-v23-1-2022-03-15.pdf/%40%40download"
    },
    {
      "title": "Financial Conflict of Interest",
      "url": "https://www.grants.nih.gov/policy-and-compliance/policy-topics/fcoi"
    },
    {
      "title": "Federal Regulations for Clinical Investigators",
      "url": "https://www.fda.gov/drugs/investigational-new-drug-application-ind/federal-regulations-clinical-investigators"
    },
    {
      "title": "E6(R3) Good Clinical Practice (GCP)",
      "url": "https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e6r3-good-clinical-practice-gcp?trk=article-ssr-frontend-pulse_little-text-block"
    },
    {
      "title": "Investigational New Drug Application (IND)",
      "url": "https://www.fda.gov/drugs/types-applications/investigational-new-drug-application-ind"
    },
    {
      "title": "IND Application Procedures: Overview",
      "url": "https://www.fda.gov/drugs/investigational-new-drug-application-ind/ind-application-procedures-overview"
    },
    {
      "title": "Electronic Source Data in Clinical Investigations",
      "url": "https://www.fda.gov/media/85183/download"
    },
    {
      "title": "Electronic Systems, Electronic Records, and Electronic Signatures in Clinical Investigations: Questions and Answers",
      "url": "https://www.fda.gov/media/166215/download?trk=article-ssr-frontend-pulse_little-text-block"
    },
    {
      "title": "ICH E8(R1) Guideline",
      "url": "https://database.ich.org/sites/default/files/E8-R1_Guideline_Step4_2021_1006.pdf"
    },
    {
      "title": "WHO data set",
      "url": "https://www.who.int/tools/clinical-trials-registry-platform/network/who-data-set"
    },
    {
      "title": "ICTRP search portal",
      "url": "https://www.who.int/tools/clinical-trials-registry-platform/the-ictrp-search-portal"
    },
    {
      "title": "Unambiguous trial identification",
      "url": "https://www.who.int/tools/clinical-trials-registry-platform/unambiguous-trial-identification"
    },
    {
      "title": "What Study Information Should Be Included on a ClinicalTrials.gov Entry?",
      "url": "https://grants.nih.gov/news-events/nih-extramural-nexus-news/2026/06/what-study-information-should-be-included-on-a-clinicaltrialsgov-entry"
    },
    {
      "title": "Clinical data publication",
      "url": "https://www.ema.europa.eu/en/human-regulatory-overview/marketing-authorisation/clinical-data-publication"
    },
    {
      "title": "Joint statement on public disclosure of results from clinical trials",
      "url": "https://www.who.int/news/item/18-05-2017-joint-statement-on-registration"
    },
    {
      "title": "Data Sharing Statements for Clinical Trials: A Requirement of the International Committee of Medical Journal Editors",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC5459581/"
    },
    {
      "title": "Chapter 15: Interpreting results and drawing conclusions",
      "url": "https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-15"
    },
    {
      "title": "Chapter 14: Completing ‘Summary of findings’ tables and grading the certainty of the evidence",
      "url": "https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-14"
    },
    {
      "title": "Chapter 10: Analysing data and undertaking meta-analyses",
      "url": "https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-10"
    },
    {
      "title": "FDA issues draft guidance regarding confirmatory evidence of clinical trials",
      "url": "https://www.fda.gov/drugs/drug-alerts-and-statements/fda-issues-draft-guidance-regarding-confirmatory-evidence-clinical-trials"
    },
    {
      "title": "Conducting Clinical Trials With Decentralized Elements: Guidance for Industry, Investigators, and Other Interested Parties",
      "url": "https://www.fda.gov/media/167696/download"
    },
    {
      "title": "Digital Health Technologies for Remote Data Acquisition in Clinical Investigations",
      "url": "https://www.fda.gov/media/155022/download"
    },
    {
      "title": "Real-World Data: Assessing Electronic Health Records and Medical Claims Data to Support Regulatory Decision-Making for Drug and Biological Products",
      "url": "https://www.fda.gov/media/152503/download"
    },
    {
      "title": "Real-World Data: Assessing Electronic Health Records and Medical Claims Data to Support Regulatory Decision-Making for Drug and Biological Products",
      "url": "https://www.hhs.gov/guidance/sites/default/files/hhs-guidance-documents/FDA/54336794fnl_RWD-Assessing-EHRs_Jan-2025-update.pdf"
    },
    {
      "title": "Low-cost dexamethasone reduces death by up to one third in hospitalised patients with severe respiratory complications of COVID-19",
      "url": "https://www.recoverytrial.net/news/low-cost-dexamethasone-reduces-death-by-up-to-one-third-in-hospitalised-patients-with-severe-respiratory-complications-of-covid-19"
    },
    {
      "title": "Making trials part of good clinical care: lessons from the RECOVERY trial",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8285150/"
    },
    {
      "title": "Guidance for best practices for clinical trials",
      "url": "https://www.who.int/our-work/science-division/research-for-health/implementation-of-the-resolution-on-clinical-trials/guidance-for-best-practices-for-clinical-trials"
    },
    {
      "title": "Guidance for best practices for clinical trials",
      "url": "https://iris.who.int/bitstream/handle/10665/378782/9789240097711-eng.pdf"
    }
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}