Skip to content
ContentLora

    Tip: press / anywhere to search.

    Explainer

    How epidemiology studies work: cohorts, case-controls and confounding

    Epidemiology studies the patterns, causes and control of disease in groups of people, often by observing them without giving any treatment.[1][2] The main observational designs are cohort, case-control and cross-sectional studies.[3] They can reveal associations, but confounding means an association may not be a cause, so evidence rankings put randomised trials above them.[4][5]

    Editor reviewedStrict sourcingUpdated Explainer18 claims · 100% primaryHealth and medicine, Science
    Reading level: 1 passage differs; the rest is the same for everyone.

    This page explains research methods. It is general information, not medical advice. A single study rarely settles a health question, so talk to a clinician before changing treatment or lifestyle because of a headline.

    Epidemiology is the study of the patterns, causes and control of disease in groups of people.[1] Much of it is observational, meaning researchers watch what happens and measure outcomes without assigning a treatment.[2]

    The main study designs

    The STROBE initiative, an international collaboration of epidemiologists, methodologists, statisticians, researchers and journal editors that sets reporting standards for this research, names three main analytical designs: cohort, case-control and cross-sectional studies.[3]

    DesignWhat it doesTypical use
    CohortFollows groups that are alike in many ways but differ in one characteristic, and compares an outcome[6]Female nurses who smoke compared with those who do not, for lung cancer[6]
    Case-controlCompares people with a disease (cases) with similar people without it (controls), looking back at their histories[7]Looking for lifestyle or exposure factors linked to a disease[7]
    Cross-sectionalCollects information at one point in time or over a short period[8]Suggesting an association between a risk factor and a disease, without proving a link[8]

    NCI notes that relative risk is used in forward-looking studies such as cohort studies and clinical trials.[9] Incidence, the number of new cases diagnosed each year, is one of the basic measures these studies track.[10] For how to read the resulting risk figures, see relative vs absolute risk.

    Bias and confounding

    In research, bias is a flaw in the design or in how information is collected or interpreted, and it can lead to wrong conclusions.[11] The Cochrane Handbook explains that it occurs when there are common causes of both the exposure or treatment people get and the outcome, so the association in the data differs from the causal effect.[4]

    A confounder is something linked to both the exposure being studied and the outcome, so a study that ties some other habit to lung cancer has to separate that habit’s effect from smoking’s.[4] NCI’s lung cancer prevention summary warns that epidemiological analyses cannot fully control for confounding by smoking.[12]

    The GRADE system, used by Cochrane to rate evidence, generally rates a body of evidence from non-randomised studies down by two levels because of the inherent risk of confounding and selection bias.[13] It can rate such evidence back up for a large effect, a dose-response gradient, or when any plausible confounding would only shrink the effect that was seen.[14] STROBE itself is a reporting checklist, and its authors say it is not a tool for judging a study’s quality.[3]

    Where observational evidence ranks

    NCI’s PDQ boards rank study designs from strongest to weakest: randomised controlled trials, then nonrandomised controlled trials, then cohort or case-control studies, then ecologic and descriptive studies, with expert opinion last.[5] The same boards note that trials are not always practical for every question, and that other designs can sometimes give solid evidence.[15] GRADE’s two-level penalty for non-randomised evidence reflects the same point: without randomisation, confounding and selection bias remain possible.[13] The site’s explainer on how aging interventions are tested shows how this plays out in one field.

    When trials overturned expectations

    Two cases show randomised trials contradicting earlier expectations.

    • Hormone therapy and heart disease. The Women’s Health Initiative ran a randomised trial of more than 68,000 postmenopausal women alongside an observational study of more than 93,000 women. The WHI found that hormone replacement therapy did not prevent heart disease as had once been thought.[16]
    • Beta-carotene and lung cancer. Two randomised trials, ATBC and CARET, found that heavy smokers taking beta-carotene supplements had higher lung cancer incidence and mortality. CARET was stopped early because its results confirmed the harm.[17]

    Gathering all the studies on a question is the job of systematic reviews.[18]

    Questions readers ask

    What is an observational study?

    It is a study in which researchers observe people or measure outcomes without trying to affect the outcome, for example without giving a treatment.[2]

    What is the difference between a cohort study and a case-control study?

    A cohort study compares an outcome in groups that are alike in many ways but differ in one characteristic, such as smoking. A case-control study starts with people who have a disease and a similar group who do not, then looks back at their histories for factors linked to the disease.[6][7]

    What is confounding?

    The Cochrane Handbook says confounding happens when something causes both the exposure or treatment people get and the outcome, so the association seen in the data differs from the true causal effect.[4]

    Can observational studies prove cause and effect?

    Not on their own. NCI notes a cross-sectional study can show an association but does not prove a link, and evidence rankings place cohort and case-control studies below randomised trials. Under GRADE, evidence from non-randomised studies is generally rated down two levels.[8][5][13]

    Have trials ever contradicted what was once believed?

    Yes. The Women's Health Initiative trial found that hormone replacement therapy did not prevent heart disease in postmenopausal women as had once been thought, and two trials found beta-carotene supplements increased lung cancer in heavy smokers.[16][17]

    Sources

    · 18 claims from 6 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]

      NCI defines epidemiology as the study of the patterns, causes and control of disease in groups of people. confirmedas of 2026-11-07

    2. [2]

      In an observational study, researchers observe people or measure outcomes without trying to affect the outcome, for example without giving a treatment. confirmedas of 2026-11-07

    3. [3]

      STROBE is an international initiative of epidemiologists, statisticians and journal editors that sets out a reporting checklist for the three main observational designs (cohort, case-control and cross-sectional), and says the checklist is not a tool for judging study quality. confirmedas of 2026-11-07

      • STROBE statement · STROBE · Primary source · What is STROBE? (retrieved 2026-11-07)
      • STROBE statement · STROBE · Primary source · Aims and use of STROBE (retrieved 2026-11-07)
      • STROBE statement · STROBE · Primary source · Aims and use of STROBE (retrieved 2026-11-07)
    4. [4]

      The Cochrane Handbook says confounding occurs when there are common causes of both the choice of intervention and the outcome, so that the observed association differs from the causal effect. confirmedas of 2026-11-07

    5. [5]

      NCI's PDQ evidence ranking puts randomized controlled trials first, then nonrandomized controlled trials, then cohort or case-control studies, then ecologic and descriptive studies, and expert opinion last. confirmedas of 2026-11-07

    6. [6]

      A cohort study compares an outcome, such as lung cancer, in groups of people who are alike in many ways but differ by one characteristic; NCI's example is female nurses who smoke compared with those who do not. confirmedas of 2026-11-07

    7. [7]

      A case-control study compares people who have a disease (cases) with a similar group who do not (controls) and looks back at their medical and lifestyle histories for factors linked to the disease. confirmedas of 2026-11-07

    8. [8]

      A cross-sectional study collects information at a single point or over a short period; it can suggest an association between a risk factor and a disease but does not prove they are linked. confirmedas of 2026-11-07

    9. [9]

      NCI notes that relative risk is used in forward-looking studies such as cohort studies and clinical trials. confirmedas of 2026-11-07

    10. [10]

      Incidence is the number of new cases of a disease diagnosed each year. confirmedas of 2026-11-07

    11. [11]

      In research, bias is a flaw in study design or in how information is collected or interpreted, and it can lead to incorrect conclusions. confirmedas of 2026-11-07

    12. [12]

      NCI's lung cancer prevention summary warns that epidemiological analyses cannot fully control for confounding by smoking. confirmedas of 2026-11-07

    13. [13]

      Under the GRADE approach used by Cochrane, evidence from non-randomized studies should generally be rated down two levels because of the risk of confounding and selection bias. confirmedas of 2026-11-07

    14. [14]

      GRADE allows certainty in non-randomized evidence to be rated up for a large effect, a dose-response gradient, or when plausible confounding would only weaken the observed effect. confirmedas of 2026-11-07

    15. [15]

      NCI's PDQ board says randomized trials are not always practical for every screening and prevention question, and that other designs can sometimes give solid evidence. confirmedas of 2026-11-07

    16. [16]

      The Women's Health Initiative ran a randomized trial of more than 68,000 postmenopausal women alongside an observational study of more than 93,000, and found that hormone replacement therapy did not prevent heart disease as had once been thought. confirmedas of 2026-11-07

    17. [17]

      Two randomized trials (ATBC and CARET) found that heavy smokers given beta-carotene supplements had higher lung cancer incidence and mortality, and CARET was stopped early for harm. confirmedas of 2026-11-07

    18. [18]

      A systematic review tries to collect all the empirical evidence that meets pre-specified eligibility criteria to answer a specific research question, using explicit methods chosen to minimise bias. confirmedas of 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: Nov 7, 2027.

    Cite this page

    "How epidemiology studies work: cohorts, case-controls and confounding." ContentLora, updated Nov 7, 2026. https://contentlora.com/explain/how-epidemiology-studies-work

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