Cohort Studies

🏛️ What Is a Cohort Study?

●​ Definition: An observational study where disease-free individuals are followed over
time to assess disease incidence based on exposure status
●​ Researcher does not assign exposure (unlike in randomized trials)
●​ Functions as a “natural experiment”
●​ Observational studies are considered “natural”

👥 Basic Structure
None
Start with: EXPOSED group vs. UNEXPOSED group
Follow over time → Track new cases of disease

●​ Source Population: Group from which exposed and unexposed are drawn
●​ Comparison: Incidence of disease in each group

🧪 Key Examples
●​ Framingham Heart Study (since 1948): Examines cardiovascular risk over generations
●​ Nurses’ Health Study (1976–present): One of the largest women's health studies

🔄 Multiple Exposures/Outcomes
●​ Cohort studies allow tracking of multiple exposures (e.g., smoking, obesity) and
multiple outcomes (e.g., cancer, CVD, death)

⏳ Types of Cohort Studies by Timing

 Type Characteristics

✔️ Less bias, better exposure

❌
Prospective Data collected in real time moving forward
data Expensive, time-consuming

Retrospective Use past records and follow up to the present ✔️ Faster, cheaper ❌ May
have poor-quality data

Ambidirectional Combines both methods

🧾 Timing Comparison Summary

Consideration Prospective Retrospective

Cost/time efficiency ❌ ✔️
Good for long-latency disease ❌ ✔️
Good exposure/confounder ✔️ ❌/✔️
data

🗂️ Sources of Exposure Info

Source Type Pros Cons

Records Inexpensive, pre-disease May lack detail/confounders

Questionnaire Good for lifestyle info Recall bias, expensive

s

Direct testing Accurate for certain Costly, impractical for large studies
exposures

🔁 Follow-Up in Cohorts
●​ Follow all participants until:
○​ Outcome develops
○​ Death
○​ Study ends
●​ Loss to follow-up (LTF) can reduce validity and sample size
📌 Strategies to Reduce LTF
●​ Gather accurate contact info (name, DOB, address, SSN)
●​ Maintain regular contact (mail, phone, social media)
●​ Offer incentives (money, calendars, etc.)

📊 Analysis in Cohort Studies

1.​ Descriptive: Are groups comparable at baseline?
2.​ Measures of Frequency & Association:
○​ Cumulative Incidence (CI) or Incidence Rate (IR)
i.​ Incidence rates (aka incidence density or person-time rates):
○​ Risk Difference (RD)​
Relative Risk (RR)

🚫 Not Ideal for Rare Diseases

To detect 500 cases over 5 years, needed sample size:

Disease Participants
Rate Needed

2/100 PY 5,000

2/10,000 PY 500,000

2/100,000 PY 5,000,000

✅ Cohort Study: Summary

🔷 Advantages
●​ Strong observational evidence
●​ Clarifies temporal relationships (exposure before disease)
●​ Good exposure/confounder info
●​ Can assess multiple outcomes/exposures

🔶 Disadvantages
●​ Costly & time-intensive
●​ Large sample often needed
●​ Poor for rare diseases
●​ Susceptible to loss to follow-up
●​ Risk of confounding​
Use stratification or multivariate analysis
○​ Confounder control only as good as its measurement
Experimental Studies

🧪 What Are Experimental Studies?

●​ Researcher assigns exposure (treatment) to participants
●​ Considered the gold standard of study design
●​ Example: Randomized Controlled Trials (RCTs)
●​ Purpose: determine cause-effect relationships

🔬 Types of Analytic (Scientific) Studies

Experimental (Intervention) Observational

Clinical Trial Cohort Study

Case-Control

🧱 Study Design Hierarchy of Evidence

From strongest to weakest:

1.​ Meta-analyses
2.​ Experimental studies (RCTs)
3.​ Cohort and case-control
4.​ Cross-sectional and ecological
5.​ Case reports, expert opinion

🔄 Experimental Research Process

1.​ Form research question
2.​ Choose study type (e.g., prevention or therapeutic trial)
3.​ Define eligibility criteria and select population
4.​ Informed consent, enrollment, random assignment
5.​ Apply strategies to minimize bias
6.​ Conduct data analysis​
⚖️ Ethical Requirement: Equipoise
●​ Investigators must genuinely not know whether treatment is better than control
●​ Study is only ethical if withholding treatment won't cause harm

📊 Types of Experimental Studies

🔹 By Purpose
Type Purpose Example

Prevention Prevent disease in healthy/high-risk Lead removal to prevent

Trial individuals poisoning
-​ Often test interventions that involve
reducing or removing a suspected
causal factor for disease

Therapeutic Treat/cure existing disease Herceptin for breast cancer

Trial (aka -​ Conducted in clinical setting survival
clinical trial) -​ Often uses clinical outcome
measure

🔹 By Unit of Allocation
Type Description Example

Individual Treatment assigned to Hepatitis C drug trial

individuals

Community Treatment assigned to groups Water fluoridation in cities

👥 Study Population
●​ Choose people at risk of the outcome and likely to adhere to treatment
●​ Must meet inclusion/exclusion criteria
●​ Must provide informed consent
●​ Generalizability depends on how broad your sample is
●​ Need enough people to have sufficient statistical power (typically >80%)
🎲 Exposure Assignment: Randomization
●​ Each participant has equal chance to receive treatment
●​ Prevents selection bias and confounding
●​ Creates balanced groups at baseline (age, smoking, BMI, etc.)
●​ Example: Physicians' Health Study (aspirin vs. placebo in 22,000+ male doctors)

🙈 Minimizing Bias
🔹 Blinding (Masking)
Type Who is unaware of group assignment?

Single-blind Participant

Double-blind Participant & treating investigator

Triple-blind Participant, treating investigator, and outcome

assessor

❗ Blinding is not always feasible (e.g., surgery or exercise interventions)

🔹 Placebo
●​ Makes experiences comparable between groups
●​ May induce “placebo effect” (apparent improvement in control group)
●​ Not ethical or feasible in all trials

🔹 Compliance
●​ Participants must follow study protocol
●​ Non-compliance reduces differences between groups
●​ Leads to bias toward the null (i.e., weakens observed effect)

📉 Analysis
 📌 Key Concepts
●​ Randomization Check: Confirm group similarity at baseline
●​ 2x2 Table: Compare disease incidence between groups
●​ Estimate:
○​ Risk Difference (RD)
○​ Relative Risk (RR) or Incidence Rate Ratio (IRR)​

🔄 Types of Analysis
Type Pros Cons

Intention-to-Treat Preserves randomizationReal-world May dilute effect due to

view noncompliance

Efficacy Analysis Estimates true biological effect Loses benefits of randomization

Example (Physicians’ Health Study)​

IRR of MI: 0.58 → 42% risk reduction in aspirin group

⚠️ Limitations of Experimental Studies

●​ Costly and time-consuming
●​ Can't study unethical or impractical exposures (e.g., smoking)
●​ May be too short to detect long-latency effects (e.g., diet & cancer)
●​ Ethical barriers: informed consent, equipoise

📝 Summary
✔ Experimental studies = most rigorous design​
✔ Randomization prevents confounding​
✔ Blinding, placebos, and compliance reduce bias​

❌
✔ Useful for prevention and treatment research​
But: Not always ethical, feasible, or affordable
Case-Control Studies

🎯 Learning Objectives
●​ Understand key components of analytic studies
●​ Identify experimental, cohort, and case-control designs
●​ Learn the strengths and limitations of case-control studies
●​ Understand how to measure and interpret odds ratios (OR)

🔍 What Is a Case-Control Study?

●​ Definition: A study where people with a disease (cases) and people without it
(controls) are selected, and their prior exposure is compared
●​ Direction: Looks backward in time from outcome to exposure
●​ Purpose: Determine whether exposure is associated with disease

🧠 Design Overview
None
Source Population
↓
Cases (with disease) ← Compare past exposures → Controls (without
disease)

●​ Controls must come from the same source population as the cases
●​ Controls must be selected independently of exposure

✅ Why Use a Case-Control Study?

Best when:

●​ Exposure data is costly or hard to obtain

●​ Disease has a long latency period (e.g., cancer)
●​ Disease is rare
 ●​ There is high loss to follow-up in a cohort
●​ You want to explore multiple exposures for a single disease

🧪 Case-Control Example:
●​ Nurses’ Health Study (blood samples from 89,949 women)
●​ Rather than test all, test:
○​ All 1,439 breast cancer cases
○​ A random sample of 2,878 controls
●​ More cost-effective, still yields valid odds ratio

🎯 Key Steps in Case-Control Studies

1.​ Define source population
2.​ Define and identify cases
○​ Use valid and accurate case definitions
3.​ Identify controls
○​ Must represent the exposure distribution of the source population
○​ Apply the “Would Criterion”: Would they have been identified as a case if they
developed the disease?

👥 Types of Controls
Type Description

Nested controls Drawn from an existing cohort study

Population-based From the same geographic region

Hospital/clinic-based From patients with unrelated illnesses

❗ Avoid diseases related to the exposure of interest when selecting hospital

controls.

📊 Measures of Association: Odds Ratio (OR)

 ❌ You Cannot Calculate:
●​ Risk
●​ Rate
●​ Risk Difference

✅ You Can Calculate:

●​ Odds Ratio (OR) – the only valid measure of association in case-control studies

➗ What Are Odds?

●​ Example: If 20 exposed people develop disease and 60 don’t, odds = 20/60 = 0.33

📐 2x2 Table Setup

Exposed Unexposed Total

Cases a b a+b

Controls c d c+d

🔁 Odds Ratio Formula

●​ OR > 1 → Exposure associated with increased odds of disease

●​ OR < 1 → Exposure associated with decreased odds
●​ OR = 1 → No association

📊 Example Calculation
 ●​ a = 20, b = 30, c = 30, d = 60

●​ Interpretation: Exposed individuals have 1.33 times the odds (33% higher odds) of
disease compared to unexposed

✅ Advantages of Case-Control Studies

●​ Less time and cost than cohort or RCT
●​ Good for rare diseases
●​ Efficient (efficiency)when exposure data is expensive
●​ Can study multiple exposures
●​ Useful in outbreak investigations

❌ Limitations of Case-Control Studies

●​ Temporal ambiguity: exposure may not precede disease
●​ Recall bias: cases may remember exposures differently than controls
●​ Selection bias: improper control selection
●​ Inefficient for rare exposures
●​ Cannot calculate risk or rate