About This Page
This is a clinician-written, evidence-based summary aligned to the USMLE Step 2 CK Content Outline. It is intended for medical students preparing for USMLE Step 2 CK. Management reflects current ACC/AHA, USPSTF, and APA guidelines. Always cross-reference with UpToDate, institutional protocols, and clinical judgment.
The Bottom Line
- RCT: investigator assigns exposure/intervention; strongest design for causality when ethical and feasible
- Cohort: start with exposure and follow to outcome; calculates incidence and relative risk
- Case-control: start with outcome and look back for exposure; best for rare diseases; calculates odds ratio
- Cross-sectional: exposure and outcome measured at one time; estimates prevalence; cannot establish temporality
- Meta-analysis statistically pools studies; quality depends on included studies, heterogeneity, and publication bias
Overview
Study design questions are among the most predictable USMLE biostatistics items. The key is to identify where the study begins: with intervention assignment, exposure status, disease status, or a population snapshot. RCTs assign treatment and follow forward; cohort studies identify exposed/unexposed groups and follow for outcomes; case-control studies identify cases/controls and look backward for exposure; cross-sectional studies measure exposure and outcome simultaneously. The design determines which measures are valid and which biases are most likely.
Epidemiology
Study designs balance internal validity, feasibility, cost, ethics, duration, and generalizability. Rare diseases are often studied with case-control designs because waiting for disease incidence in a cohort would be inefficient. Rare exposures are often studied with cohort designs because investigators can deliberately assemble exposed and unexposed groups. RCTs minimize confounding through randomization but may have limited external validity because trial patients are selected and monitored more intensively than real-world patients.
Study Design Recognition
Symptoms
Investigator randomly assigns drug, vaccine, screening intervention, or placebo and follows outcomes = randomized controlled trial
Groups are defined by exposure and followed over time to see who develops disease = cohort study
Groups are defined by disease status and investigators look backward for exposure = case-control study
Exposure and outcome measured at a single point in time = cross-sectional study
Population-level exposure compared with population-level outcome = ecological study; beware ecological fallacy
Descriptive report of unusual patients without control group = case series or case report
Signs
Relative risk requires incidence data and is typical for cohort studies and RCTs
Odds ratio is classic for case-control studies and approximates relative risk when disease is rare
Prevalence ratio or prevalence odds ratio may arise from cross-sectional studies
Randomization reduces confounding; blinding reduces performance and ascertainment bias
Allocation concealment prevents selection bias before randomization assignment is known
Measures by Study Design
First-line
Randomized controlled trialMeasures incidence, relative risk, absolute risk reduction, number needed to treat, and adverse event rates. Best for intervention efficacy
Cohort studyProspective or retrospective. Measures incidence, relative risk, attributable risk. Best for rare exposures and multiple outcomes from one exposure
Case-control studyMeasures odds ratio. Efficient for rare diseases or long latency; cannot directly calculate incidence or relative risk from sampling scheme
Cross-sectional studyMeasures prevalence. Useful for disease burden and associations, but temporality is unclear
Second-line
Systematic reviewStructured, reproducible literature review with explicit inclusion/exclusion criteria and quality assessment
Meta-analysisQuantitative pooling of results; forest plots show effect estimates and confidence intervals; heterogeneity affects interpretation
Ecological studyUses group-level data. Useful for policy or environmental exposures but vulnerable to ecological fallacy
Specialist
Crossover trialEach participant receives multiple interventions in sequence, serving as their own control; requires washout and stable condition
Noninferiority trialTests whether new treatment is not worse than standard by more than a prespecified margin; common when new treatment is safer, cheaper, or easier
How to Select or Interpret Study Designs
Standard epidemiology and evidence-based medicine principles used in USMLE Step 2 CK1
Best design by research question
- Therapeutic efficacy: randomized controlled trial if ethical and feasible
- Harmful exposure when randomization is unethical: cohort or case-control depending on frequency and latency
- Rare disease: case-control study
- Rare exposure: cohort study
- Prevalence/burden: cross-sectional study
2
Common measures
- RCT/cohort: relative risk = risk in exposed / risk in unexposed
- Case-control: odds ratio = odds of exposure among cases / odds of exposure among controls
- Cross-sectional: prevalence = existing cases / population at a point or period
- Meta-analysis: pooled effect estimate with confidence interval; assess heterogeneity and publication bias
3
Validity features
- Randomization balances known and unknown confounders on average
- Blinding reduces behavior and measurement differences caused by knowledge of group assignment
- Intention-to-treat preserves benefits of randomization by analyzing participants in assigned groups
- Loss to follow-up threatens validity, especially if differential between groups
4
Limitations
- Observational studies can show association but are more vulnerable to confounding
- Cross-sectional studies cannot determine whether exposure preceded outcome
- Case-control studies are vulnerable to recall and selection bias
- Meta-analysis can produce misleading precision if poor-quality or heterogeneous studies are pooled
Complications
- Confounding: Exposure-outcome association distorted by a third variable related to both
- Selection bias: Study groups differ because of how participants are chosen or retained
- Recall bias: Cases remember prior exposures differently than controls, especially in case-control studies
- Loss to follow-up: Attrition distorts incidence estimates and treatment comparisons
- Publication bias: Positive studies are more likely to be published and included in meta-analysis
- Ecological fallacy: Group-level association incorrectly applied to individuals
USMLE Step 2 CK Exam Tips
- 1Case-control starts with disease; cohort starts with exposure
- 2Rare disease = case-control; rare exposure = cohort
- 3Case-control gives odds ratio, not relative risk
- 4Cohort and RCT can calculate incidence and relative risk
- 5Cross-sectional gives prevalence and cannot prove temporality
- 6Randomization reduces confounding; blinding reduces measurement/behavior bias
- 7Intention-to-treat analyzes patients in their assigned groups even if they did not comply
- 8Meta-analysis is only as good as the quality and heterogeneity of included studies
practicetest your knowledge on biostatistics — study designsApply what you've learnt with USMLE Step 2 CK-style questions from the iatroX Q-Bank — preventive medicine and beyond.
open q-bank