Human Subjects Research

Human subjects research presents unique challenges compared to basic science research. Because it involves real people, it must account for ethical considerations and the complexity of human variability. Researchers face real limits on how much they can control variables — but with thoughtful design and careful methodology, they can still draw meaningful conclusions.

Key Takeaways

• The experimental approach manipulates a variable directly, using randomization, blinding, and statistical adjustment to isolate its effect.

• The observational approach examines exposures and outcomes without manipulation: cohort (prospective), cross-sectional (single point in time), case-control (retrospective, good for rare outcomes).

• Hill's criteria help judge whether an observational association is likely causal — but observational findings should still be described as correlations.

• Selection bias, detection bias, observation (Hawthorne) bias, and confounding are four distinct ways human subjects research can be distorted — know the difference between an unrepresentative sample, inconsistent outcome-searching, observation-induced behavior change, and a hidden third variable.

The Experimental Approach

The experimental approach, common in biomedical research, manipulates an independent variable to determine its effect on an outcome while carefully controlling for external factors. Three elements help ensure reliable results.

Randomization

Randomization evenly distributes confounding variables between groups. In a clinical trial testing a new medication, participants are randomly assigned to the treatment or control group. This reduces the risk of unintentional bias — differences in age, gender, or pre-existing conditions — so that any observed effect is more likely to come from the treatment itself rather than unrelated factors.

Blinding (Single- vs. Double-Blind)

Blinding removes bias by limiting what subjects and researchers know about treatment groups.

• Single-blind: participants don't know whether they're receiving the treatment or a placebo, minimizing the placebo effect (perceived benefit from believing you're receiving treatment).

• Double-blind: neither participants nor researchers know who's in which group, further reducing researcher bias — where expectations could unintentionally influence data collection or interpretation.

Data Analysis

Researchers use statistical techniques — such as regression models — to adjust for potential confounders like age or baseline health differences between groups, helping ensure results are valid and meaningfully reflect the treatment's effectiveness.

The Observational Approach

The observational approach is used when experiments aren't feasible or ethical. It doesn't involve direct manipulation of variables; instead, it examines relationships between exposures and outcomes.

Observational study types

Study type	Timing	What it does	Best for
Cohort study	Prospective (forward in time)	Follows a group over time to assess outcome rate based on exposure	Establishing temporality (e.g., tracking smokers/non-smokers 20 years for lung cancer rates)
Cross-sectional study	Single point in time	Assesses exposure and outcome simultaneously	Measuring prevalence (e.g., lung cancer rate in smokers vs. nonsmokers right now)
Case-control study	Retrospective (backward in time)	Compares exposure history between those with the outcome (cases) and without it (controls)	Studying rare outcomes (e.g., past smoking habits in lung cancer patients vs. not)

Hill's Criteria for Causality

When evaluating observational studies, researchers use Hill's criteria — developed by Sir Austin Bradford Hill in 1965 — to judge whether an observed association is likely to be causal. These criteria don't provide an absolute guideline, so any relationship from an observational study should still be described as a correlation, not a proven cause.

1. Temporality — the exposure must occur before the outcome (smoking must precede lung cancer).

2. Strength — stronger associations (e.g., a much higher lung cancer risk in smokers) better support causality.

3. Dose-response relationship — as exposure increases, the outcome's likelihood or severity should also increase (heavier smoking, higher risk).

4. Consistency — the association should hold across multiple studies and populations.

5. Plausibility — there should be a biologically reasonable mechanism (carcinogens in tobacco smoke damaging lung tissue).

6. Specificity — a specific exposure should lead to a specific outcome (asbestos exposure linked closely to mesothelioma).

7. Coherence — the association should align with existing knowledge.

8. Experiment — experimental evidence, when possible, strengthens the causal link.

9. Consideration of alternative explanations — researchers must rule out other factors (e.g., genetic predisposition) that could explain the relationship.

Hill's original 1965 framework is classically nine "viewpoints," with the ninth traditionally called analogy (judging by similarity to already-accepted cause-effect relationships) rather than "consideration of alternative explanations." The version above reflects how this course teaches the ninth criterion — a common exam-prep framing that captures similar reasoning (ruling out confounders and other explanations) — so it's worth knowing both names if you encounter either on a passage.

Error Sources in Human Subjects Research

Bias types in human subjects research

Bias type	Definition	Example
Selection bias	The sample isn't representative of the population being studied	Recruiting only healthy volunteers for a chronic-illness study
Detection bias	Outcomes are searched for inconsistently across groups	Screening smokers for lung cancer more than non-smokers, overestimating prevalence in smokers
Observation bias (Hawthorne effect)	Participants change behavior because they know they're being observed	Diet-study participants eating healthier while monitored
Confounding	A third variable is associated with both the exposure and the outcome, creating a false or misleading association	Ice cream consumption appears linked to drowning, but both are driven by hot weather

Common MCAT Mistakes

• Assuming an observational study can prove causation. Even when all of Hill's criteria are satisfied, an observational association is still a correlation, not proof — only well-controlled experiments can establish causality directly.

• Mixing up cohort, cross-sectional, and case-control studies. Cohort studies move forward in time from exposure to outcome, case-control studies move backward from outcome to exposure, and cross-sectional studies capture both at a single moment — confusing the timing direction misreads the whole study design.

• Confusing single-blind and double-blind designs. Single-blind hides group assignment from participants only (controlling for the placebo effect); double-blind hides it from participants and researchers (also controlling for researcher bias). Assuming blinding always means "double" overlooks a key distinction.

• Mistaking the Hawthorne effect for confounding. The Hawthorne effect is a behavior change caused by being observed; confounding is a hidden third variable linked to both exposure and outcome. They distort results in different ways and call for different fixes.

MCAT-Style Concept Check

Question: A study finds that people who carry lighters have a higher rate of lung cancer than people who don't. The researchers conclude that carrying a lighter increases lung cancer risk. Which source of error best explains why this conclusion is flawed?

• A) Detection bias, because lung cancer is diagnosed more often in lighter carriers

• B) The Hawthorne effect, because participants changed their behavior once observed

• C) Confounding, because smoking is linked to both carrying a lighter and lung cancer

• D) Selection bias, because the sample of lighter carriers wasn't representative

Answer: C

Explanation: Carrying a lighter and having lung cancer are both driven by a third variable — smoking — rather than one causing the other. This is a textbook case of confounding: a variable associated with both the "exposure" (carrying a lighter) and the outcome (lung cancer) creates a misleading association, much like the ice cream–drowning example tied to hot weather. Detection bias would require inconsistent screening between groups, the Hawthorne effect requires participants knowingly changing behavior under observation, and selection bias would require an unrepresentative sample — none of which is described here.

FAQ

What's the difference between the experimental and observational approach?

The experimental approach directly manipulates an independent variable and controls for outside factors using randomization, blinding, and statistical adjustment. The observational approach doesn't manipulate anything — it examines relationships between exposures and outcomes that already exist, which is used when direct manipulation isn't feasible or ethical.

What's the difference between single-blind and double-blind studies?

In a single-blind study, only the participants don't know their group assignment, which limits the placebo effect. In a double-blind study, neither participants nor researchers know the assignments, which also limits researcher bias.

How do cohort, cross-sectional, and case-control studies differ?

A cohort study follows a group forward in time from exposure to outcome. A cross-sectional study measures exposure and outcome at a single point in time. A case-control study works backward, comparing exposure history between people who already have the outcome and those who don't — useful for studying rare outcomes.

Can Hill's criteria prove that an association is causal?

No. Hill's criteria — including temporality, strength, dose-response, consistency, and plausibility — help judge how likely an observational association is to be causal, but they don't provide an absolute guideline. Findings from observational studies should still be described as correlations, not proven causes.