Professional Role · Topic 19 of 21

Research Design & Research Ethics

Professional role · research design and ethics

Research Design & Research Ethics

Randomize to defeat confounding, follow cohorts forward, run case-control studies backward — and protect every subject through the IRB, the Belmont principles and voluntary consent.

Key numbers

IRB: at least 5 members, including unaffiliated≥ 5 members

01 · The ruleBelmont: respect for persons, beneficence, justice

The Belmont Report's three principles for human-subjects research are respect for persons (applied through informed consent), beneficence (risk–benefit assessment) and justice (fair selection of subjects).

02 · What it isRCT: randomization balances known and unknown confounders

A randomized controlled trial assigns the intervention by chance, making groups comparable in measured and unmeasured factors, so outcome differences can be attributed to the intervention — the strongest single design for treatment questions.

03 · Don't confuseCohort follows exposure; case-control starts from outcome

A cohort study follows exposed and unexposed people forward to see who develops the outcome (incidence, relative risk); a case-control study starts from cases and matched controls and looks back at exposures (odds ratio).

The rules

Know these

IRB approval before human-subjects research must knowHuman-subjects research — a systematic investigation designed to contribute to generalizable knowledge — needs IRB approval before it starts; the IRB checks that risks are minimized and reasonable, selection is equitable and consent adequate.

Research consent: voluntary, withdraw anytime without penalty must knowResearch consent opens with the key information needed to decide, and must state that participation is voluntary, refusal brings no penalty or loss of benefits, and the subject may withdraw at any time without penalty.

Case-control: best for rare diseases; watch recall biasCase-control studies are ideal for rare diseases and are cheap and fast, but cases recall past exposures more completely than controls (recall bias) and poorly chosen controls introduce selection bias.

Cohort: establishes temporality; loses subjects over timeBecause cohort members are free of the outcome at entry, a cohort establishes that exposure preceded the outcome and gives incidence and relative risk; long follow-up brings loss to follow-up, cost and unmeasured confounding.

Allocation concealment and blinding prevent biasAllocation concealment keeps anyone from knowing the next assignment, preventing biased allocation; blinding (often with a matching placebo) keeps participants, clinicians and outcome assessors from shading results toward the expected effect.

Extra protections: pregnancy, prisoners, childrenFederal rules add protections for pregnant women, fetuses and neonates (Subpart B), prisoners (Subpart C) and children (Subpart D), and require added safeguards for adults with impaired decision-making capacity or economic or educational disadvantage.

Children: parental permission plus the child's assentResearch with children needs a parent's or guardian's permission plus the child's assent — an affirmative agreement; mere failure to object is not assent.

IRB: at least 5 members, including unaffiliatedAn IRB has at least five members of varied backgrounds, including at least one scientist, one nonscientist and one person not affiliated with the institution; members with a conflicting interest cannot review that project.

What it is

Core idea

Cross-sectional study: a prevalence snapshotA cross-sectional study measures exposure and outcome at one point in time — a snapshot that is good for prevalence and quick and cheap, but cannot show which came first.

Observational versus experimental: who assigns exposureIf the investigator assigns the exposure, the study is experimental (interventional); if exposure occurs naturally and is only observed (cohort, case-control, cross-sectional), it is observational. Descriptive studies describe; analytical studies test associations.

Intention-to-treat: analyze as randomizedIntention-to-treat analysis keeps every participant in the group they were randomized to, whatever treatment they actually received; per-protocol analysis includes only adherent participants. Superiority trials should always report intention-to-treat.

Selection bias, information bias, confoundingSelection bias: the sample differs from the population it represents. Information bias: exposure or outcome is mismeasured, as in recall bias. Confounding: a third factor linked to both exposure and outcome distorts the association.

Cohort or case-control?

Commonly confused

Starts from	Exposure — people free of the outcome
Direction	Forward in time (usually prospective)
Measure of association	Incidence and relative risk
Best for	Showing exposure came before outcome; several outcomes
Main weakness	Loss to follow-up, cost, confounding

Commonly confused

Tell apart

Internal validity versus external validityInternal validity asks whether the design, conduct and analysis answer the question without bias; external validity asks whether the findings generalize to other patients and settings (ecological validity: to real-life practice).

Coercion versus undue influenceCoercion is an overt threat of harm to obtain compliance; undue influence is an excessive, unwarranted or improper reward or offer. Either invalidates research consent, which must be voluntary.

More detail

Reference

Show 4 reference facts

Minimal risk means the probability and magnitude of harm are no greater than those of daily life or routine physical or psychological examinations.

In a crossover trial each participant receives both interventions in sequence, usually in random order, serving as their own control.

Case reports and case series describe one or a few patients — useful for new findings and hypotheses (e.g., a new drug adverse effect) but at risk of reflecting chance.

The Belmont Report cites the Tuskegee syphilis study — effective treatment withheld from disadvantaged Black men — as a violation of justice in the selection of research subjects.

Pearl

Belmont maps one-to-one: respect for persons → informed consent; beneficence → risk–benefit review; justice → fair selection of subjects.

Sources · Ranganathan P, Aggarwal R. Study designs: Part 1 — An overview and classification. Perspect Clin Res 2018;9:184–186 · Aggarwal R, Ranganathan P. Study designs: Part 2 — Descriptive studies. Perspect Clin Res 2019;10:34–36 · Ranganathan P, Aggarwal R. Study designs: Part 3 — Analytical observational studies. Perspect Clin Res 2019;10:91–94 · Aggarwal R, Ranganathan P. Study designs: Part 4 — Interventional studies. Perspect Clin Res 2019;10:137–139 · Aggarwal R, Ranganathan P. Study designs: Part 5 — Interventional studies (II). Perspect Clin Res 2019;10:183–186 · Ranganathan P, Aggarwal R. Study designs: Part 5 — Interventional studies (III). Perspect Clin Res 2020;11:47–50 · Andrade C. Internal, external, and ecological validity in research design, conduct, and evaluation. Indian J Psychol Med 2018;40:498–499 · HHS Office for Human Research Protections — The Belmont Report (1979) · 45 CFR Part 46 — Protection of Human Subjects (Common Rule and Subparts B–D), eCFR · Starter sheet built from the verified AGACNP fact map (facts reviewed 2026-10-15).