Skip to content
ContentLora

    Tip: press / anywhere to search.

    Explainer

    How clinical trials work: phases, randomisation and reading results

    Clinical trials test a treatment in people under a written protocol, and they are the main way researchers find out whether a new treatment is safe and effective.[1][2] Drugs move from small Phase 1 safety studies to Phase 3 trials of 300 to 3,000 people before approval, and only a minority make it through every phase.[3][4][5] Randomisation, placebos and blinding reduce bias, and public registries let anyone check what a trial set out to measure.[6][7][8]

    Editor reviewedStrict sourcingUpdated Explainer35 claims · 100% primaryHealth and medicine, Science
    Reading level: 1 passage differs; the rest is the same for everyone.

    This page explains how clinical trials are designed and reported. It is not medical advice. Anyone thinking about joining a trial, or about what a trial result means for their own care, should talk to a clinician.

    A clinical trial tests a medical, surgical or behavioural intervention in people, and trials are the primary way researchers find out whether a new treatment or prevention method is safe and effective.[1] For registration purposes, WHO defines a trial as any study that prospectively assigns people to a health intervention to measure its effects on health, from Phase I to Phase IV.[9] Each trial follows a protocol. It fixes in advance who can join, how many people take part, how long the study runs, whether there is a control group and how the data will be analysed.[2] Before it starts, an Institutional Review Board checks that the trial is ethical, avoids unneeded risks and has a safety plan.[10]

    The four phases

    New drugs are tested in steps, and each step involves more people.[3][5]

    • Phase 1 gives the drug to 20 to 100 people, often healthy volunteers, to study safety and dosage.[3] Cancer drugs are an exception: they are first tested in patients with that cancer.[11]
    • Phase 2 treats up to several hundred patients. It is usually too small to prove the drug helps, but it adds safety data and shapes the next trial.[4][12]
    • Phase 3, the “pivotal” stage, enrols 300 to 3,000 patients for 1 to 4 years to show whether the drug offers a treatment benefit.[5]
    • Phase 4 happens after approval and watches for side effects that earlier trials missed.[13]

    Most drugs drop out along the way. FDA estimates that about 70% move on from Phase 1, about 33% from Phase 2 and 25-30% from Phase 3.[3][4][5]

    FDA’s patient guide gives typical sizes and durations: Phase 1, 20 to 100 participants over several months; Phase 2, up to several hundred over several months to 2 years; Phase 3, 300 to 3,000 over 1 to 4 years.[3][4][5] Phase 3 supplies most of the safety data because size and duration make rare or long-term adverse effects more visible.[14] After Phase 3, the sponsor files a New Drug Application with everything from preclinical data to Phase 3 results. Once FDA accepts it as complete, the review team has 6 to 10 months to decide.[15] Phase 4 studies can run to thousands of patients and look at long-term effectiveness as well as late harms.[13]

    Randomisation, placebos and blinding

    In a randomised trial, people are assigned to groups by chance. This makes the groups similar, so differences in outcome can be credited to the treatment more fairly.[6] The comparison group, or control group, does not receive the new treatment.[16] It may get a placebo, an inactive substance or procedure that looks the same and is given the same way as the real one.[17] In a double-blind trial, neither participants nor researchers know who received what until the trial ends, which makes the results less likely to be biased.[7]

    A real example shows these features together. The Phase 3 HAELO trial randomised 80 patients 2:1 to a single infusion of the gene-editing therapy lonvo-z or a placebo.[18]

    Trials versus observational studies

    Observational studies follow people in ordinary settings and do not test an intervention.[19] A case-control study, for example, compares people who have a condition with similar people who do not, looking back at their histories for factors linked to it.[20] Such studies can suggest which treatments or prevention strategies are worth testing, but the trial is the step that tests them.[19][1]

    Regulators sometimes insist on randomised outcome data. Under FDA’s 2025 COVID-19 vaccine framework, approval for healthy people requires randomised trials that measure clinical outcomes, and Pfizer-BioNTech agreed to a saline placebo-controlled trial in healthy adults aged 50-64.[21]

    Endpoints and how to read results

    The endpoint is the outcome a trial measures to judge whether the intervention helps, such as survival or symptom relief.[22] A surrogate endpoint, such as a shrinking tumour or a lower biomarker level, can be measured sooner and may allow earlier approval. It is not always a true sign of how well a treatment works.[23] Outcome trials that count heart attacks or kidney failure are covered in GLP-1 outcome trials explained, and longevity trials in how aging interventions are tested.[24]

    Results usually come as ratios.

    • Hazard ratio: compares how often an event happens over time in two groups. A value of one means no difference.[25]
    • Relative risk: compares the risk in one group with the risk in another. Again, one means no difference.[26]
    • Absolute difference: in the SELECT trial, a hazard ratio of 0.80 corresponded to event rates of 6.5% on semaglutide versus 8.0% on placebo.[24]
    • Statistical significance: a difference counts as statistically significant when it is bigger than chance alone would be expected to produce.[27]

    Trials can also stop early. In August 2026 BioNTech ended a phase 2 cancer vaccine trial after its data monitoring board saw a numerical overall-survival imbalance between arms.[28]

    Registration: checking what a trial promised

    WHO calls registering every interventional trial a scientific, ethical and moral responsibility. Its International Clinical Trials Registry Platform links registries through one search portal but is not a registry itself.[8] In the US, federal rules require applicable clinical trials started after 27 September 2007 to be registered on ClinicalTrials.gov.[29] Results are generally due within one year of the primary completion date.[30] On 7 November 2026, ClinicalTrials.gov held 610,756 study records.[31]

    Journals add their own pressure. The International Committee of Medical Journal Editors asks editors to consider trials only if they were registered before the first participant enrolled. It accepts WHO primary registries and ClinicalTrials.gov.[32][33]

    In the US, the agency that reviews trial data for drugs, biologics and devices is the us-food-and-drug-administration.[34] Vaccine recommendations after approval come from CDC and its advisory committee.[35]

    Next in the crash course

    More in Health & medicine

    Go deeper

    Questions readers ask

    What are the phases of a clinical trial?

    FDA describes Phase 1 as 20 to 100 people testing safety and dosage, Phase 2 as up to several hundred people studying efficacy and side effects, and Phase 3 as 300 to 3,000 people showing whether the product offers a benefit. Phase 4 studies follow approval and look for side effects not seen earlier.[3][4][5][13]

    What share of drugs get through the trial phases?

    FDA says about 70% of drugs move on from Phase 1, about 33% from Phase 2 and about 25-30% from Phase 3.[3][4][5]

    Why are trials randomised and blinded?

    Dividing people into groups by chance makes the groups similar, so treatments can be compared fairly. In a double-blind trial neither participants nor researchers know who gets which treatment until the end, which makes the results less likely to be biased.[6][7]

    What is the difference between a clinical trial and an observational study?

    A clinical trial assigns people to an intervention and measures its effects. An observational study watches people without testing an intervention, and can point to ideas that trials then test.[9][19]

    Do trials have to be registered?

    US rules require applicable clinical trials started after 27 September 2007 to be registered, and results are generally due within a year of the primary completion date. Medical journal editors in the ICMJE also require registration before the first participant enrols as a condition of publication.[29][30][32]

    What does a hazard ratio of 0.80 mean?

    A hazard ratio compares how often an event happens over time in two groups, and a value of one means no difference. In the SELECT trial, a hazard ratio of 0.80 matched event rates of 6.5% on semaglutide versus 8.0% on placebo.[25][24]

    Sources

    · 35 claims from 12 publishers

    Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.

    1. [1]

      The US National Institute on Aging describes clinical trials, which test a medical, surgical or behavioural intervention in people, as the primary way researchers determine whether a new treatment or prevention method is safe and effective. confirmedas of 2026-11-07

    2. [2]

      Clinical trials follow a written study plan, called a protocol, which sets who can take part, how many people, how long the study lasts, whether there is a control group, how the product is given and how the data will be analysed. confirmedas of 2026-11-07

      • Step 3: Clinical Research · US Food and Drug Administration · Primary source · Designing Clinical Trials (retrieved 2026-11-07)
      • Step 3: Clinical Research · US Food and Drug Administration · Primary source · Designing Clinical Trials (retrieved 2026-11-07)
    3. [3]

      FDA describes Phase 1 studies as involving 20 to 100 healthy volunteers or people with the condition for several months, with the purpose of testing safety and dosage; about 70% of drugs move on to the next phase. confirmedas of 2026-11-07

    4. [4]

      FDA describes Phase 2 studies as involving up to several hundred people with the condition for several months to 2 years, to study efficacy and side effects; about 33% of drugs move on to the next phase. confirmedas of 2026-11-07

    5. [5]

      FDA describes Phase 3 studies, sometimes called pivotal studies, as involving 300 to 3,000 people with the condition over 1 to 4 years, designed to show whether a product offers a treatment benefit; about 25-30% of drugs move on to the next phase. confirmedas of 2026-11-07

    6. [6]

      In a randomised trial, participants are divided into groups by chance, which makes the groups similar so that the effects of the treatments can be compared more fairly. confirmedas of 2026-11-07

    7. [7]

      In a double-blind trial, neither the participants nor the researchers know who is receiving which treatment until the trial is over, which makes the results less likely to be biased. confirmedas of 2026-11-07

    8. [8]

      WHO says registering all interventional trials is a scientific, ethical and moral responsibility; its International Clinical Trials Registry Platform is not itself a registry but links registries through a search portal. confirmedas of 2026-11-07

    9. [9]

      For registration purposes, WHO defines a clinical trial as any study that prospectively assigns people to one or more health-related interventions to evaluate the effects on health outcomes, covering Phase I to Phase IV trials. confirmedas of 2026-11-07

    10. [10]

      Before a trial begins, an Institutional Review Board reviews it to check that it is well designed, legal and ethical, avoids unneeded risks and has a safety plan. confirmedas of 2026-11-07

    11. [11]

      FDA says that when a new drug is intended for cancer patients, Phase 1 studies are run in patients with that type of cancer rather than in healthy volunteers. confirmedas of 2026-11-07

    12. [12]

      FDA says Phase 2 studies are typically not large enough to show whether a drug will be beneficial; they mainly add safety data and help design Phase 3 protocols. confirmedas of 2026-11-07

    13. [13]

      Phase 4 studies take place after a product is approved, can involve several thousand people, and look for side effects not seen in earlier trials and for long-term effectiveness. confirmedas of 2026-11-07

    14. [14]

      FDA says Phase 3 studies provide most of the safety data, and because they are larger and longer they are more likely to reveal long-term or rare side effects missed earlier. confirmedas of 2026-11-07

    15. [15]

      To market a new drug, a developer files a New Drug Application containing everything from preclinical data to Phase 3 results; once FDA accepts the application as complete, its review team has 6 to 10 months to decide whether to approve. confirmedas of 2026-11-07

      • Step 4: FDA Drug Review · US Food and Drug Administration · Primary source · New Drug Application (retrieved 2026-11-07)
      • Step 4: FDA Drug Review · US Food and Drug Administration · Primary source · FDA Review (retrieved 2026-11-07)
    16. [16]

      The control group in a clinical trial is the group that does not receive the new treatment being studied; it is compared with the group that does. confirmedas of 2026-11-07

    17. [17]

      A placebo is an inactive substance or intervention that looks the same as, and is given the same way as, the treatment being tested. confirmedas of 2026-11-07

    18. [18]

      The HAELO Phase 3 trial randomized 80 patients 2:1 to a single 50 mg infusion of lonvo-z or placebo; median follow-up was 7.5 months as of 10 February 2026, and no serious or Grade 3 or higher adverse events were reported in the lonvo-z group. confirmedas of 2026-06-13

    19. [19]

      Observational studies watch people in normal settings and do not test an intervention; they can help identify treatments or prevention strategies to test later in clinical trials. confirmedas of 2026-11-07

    20. [20]

      A case-control study compares people who have a condition with similar people who do not, looking back at their histories to find factors that may be associated with the condition. confirmedas of 2026-11-07

    21. [21]

      Under the FDA framework, approval for healthy people requires randomized trials measuring clinical outcomes; Pfizer-BioNTech agreed to a saline placebo-controlled trial in healthy adults aged 50-64. confirmedas of 2025-08-25

    22. [22]

      An endpoint is an event or outcome, such as survival or symptom relief, measured objectively to decide whether the intervention being studied is beneficial. confirmedas of 2026-11-07

    23. [23]

      A surrogate endpoint, such as a shrinking tumour or a lower biomarker level, is used in place of a stronger outcome because it can be measured sooner and may allow earlier approval, but it is not always a true indicator of how well a treatment works. confirmedas of 2026-11-07

    24. [24]

      In the SELECT trial of 17,604 adults aged 45 or older with cardiovascular disease and overweight or obesity but no diabetes, semaglutide 2.4 mg reduced cardiovascular death, heart attack or stroke to 6.5% versus 8.0% on placebo (hazard ratio 0.80). confirmedas of 2023-11-11

    25. [25]

      A hazard ratio compares how often an event happens over time in one group versus another; a hazard ratio of one means no difference between the groups. confirmedas of 2026-11-07

    26. [26]

      Relative risk compares the risk of an event in one group with the risk in another; a relative risk of one means no difference. confirmedas of 2026-11-07

    27. [27]

      A difference between groups is called statistically significant when it is greater than would be expected by chance alone. confirmedas of 2026-11-07

    28. [28]

      On 28 August 2026 BioNTech ended a phase 2 trial of autogene cevumeran monotherapy in resected colorectal cancer after its data monitoring board saw a numerical overall-survival imbalance between arms; the pancreatic cancer trial IMcode003 continues. confirmedas of 2026-08-28

    29. [29]

      US federal rules (42 CFR Part 11) require registration on ClinicalTrials.gov of any "applicable clinical trial" (a category the rule defines, covering certain trials of FDA-regulated drug and device products) initiated after 27 September 2007. confirmedas of 2026-10-01

    30. [30]

      Under 42 CFR 11.44, results for applicable clinical trials must generally be submitted no later than one year after the trial's primary completion date, with delays allowed in some cases where approval of a new use is being sought. confirmedas of 2026-10-01

    31. [31]

      On 7 November 2026 the ClinicalTrials.gov public API reported 610,756 study records. confirmedas of 2026-11-07

    32. [32]

      The International Committee of Medical Journal Editors recommends that journals consider a trial for publication only if it was registered in a public registry at or before the first participant's consent to enrol. confirmedas of 2026-11-07

    33. [33]

      ICMJE accepts registration in any WHO primary registry or in ClinicalTrials.gov, which is a data provider to the WHO platform. confirmedas of 2026-11-07

    34. [34]

      FDA is responsible for protecting public health by ensuring the safety, efficacy and security of human and veterinary drugs, biological products and medical devices, and the safety of the food supply, cosmetics and radiation-emitting products. confirmedas of 2026-11-07

      • What We Do · US Food and Drug Administration · Primary source · FDA Mission (retrieved 2026-11-07)
    35. [35]

      The Advisory Committee on Immunization Practices (ACIP) is a federal advisory committee of medical and public health experts that develops recommendations on vaccine use, and CDC sets the US adult and childhood immunisation schedules based on its recommendations. confirmedas of 2025-08-12

      • About ACIP · Centers for Disease Control and Prevention · Primary source · 2025-08-12 · Read through WebFetch (cdc.gov blocks curl) (retrieved 2026-11-07)
      • About ACIP · Centers for Disease Control and Prevention · Primary source · 2025-08-12 · Read through WebFetch (cdc.gov blocks curl) (retrieved 2026-11-07)
    Revision history (1)
    1. Page created.

    Created Nov 7, 2026. Last reviewed by an editor on Nov 8, 2026. Next scheduled review: May 7, 2027.

    Cite this page

    "How clinical trials work: phases, randomisation and reading results." ContentLora, updated Nov 7, 2026. https://contentlora.com/explain/how-clinical-trials-work

    Spotted an error? Suggest a correction or emailcorrections@contentlora.com.