A researcher wants to study the causes of a rare cancer that takes 20 years to develop. Which study design is most appropriate?
ACross-sectional survey
BRandomized controlled trial
CCase-control study
DProspective cohort study
Case-control studies are ideal for rare diseases because they start with people who already have the disease (cases) and compare them to those who don't (controls), avoiding the need to follow thousands of participants for decades hoping to observe rare events. A prospective cohort would require an enormous sample and decades of follow-up. An RCT is unethical for suspected carcinogens. A cross-sectional study cannot establish temporal order.
Question 2 True / False
A case-control study can directly calculate relative risk (risk ratio) from its data, just like a cohort study.
TTrue
FFalse
Answer: False
Case-control studies work backward from outcome to exposure, so they never observe incidence rates in the exposed and unexposed populations. They can only calculate an odds ratio (the ratio of odds of exposure in cases vs. controls). The odds ratio approximates the risk ratio only when the disease is rare — the 'rare disease assumption.' Cohort studies, which follow participants forward in time, can directly calculate incidence rates and therefore relative risk.
Question 3 Short Answer
Why does random allocation in an RCT eliminate confounding, while matching or restriction in an observational study can only partially control for it?
Think about your answer, then reveal below.
Model answer: Random allocation distributes both known and unknown confounders equally across treatment groups by chance, so no systematic differences exist between groups at baseline. Matching and restriction in observational studies can only control for confounders the researcher has already identified and measured — unmeasured or unanticipated confounders remain unbalanced and can still bias the estimate.
This distinction is why RCTs are considered the gold standard for causal inference. Observational studies rely on researchers correctly anticipating every possible confounder, measuring it accurately, and adjusting for it statistically — an impossible standard. Randomization sidesteps this problem entirely by making confounding unlikely through the laws of probability.