Biostatistics, Epidemiology & EBM

In a randomized controlled trial (RCT), what single design feature makes it the strongest design for establishing causation?

Randomization of subjects to intervention vs control. Random allocation distributes both known and unknown confounders equally between groups, so a difference in outcome can be attributed to the intervention itself. RCTs are prospective and interventional.

What is the difference between a cohort study and a case-control study in terms of starting point?

A cohort study starts with EXPOSURE status (exposed vs unexposed) and follows subjects forward to see who develops the OUTCOME. A case-control study starts with OUTCOME status (cases with disease vs controls without) and looks BACKWARD to compare exposures.

Which measure of association is the appropriate output of a case-control study, and why can't it directly yield relative risk?

The odds ratio (OR). A case-control study starts by selecting on disease status, so the investigator fixes the number of cases and controls and cannot measure incidence/risk in the population — therefore relative risk cannot be computed directly. The OR approximates RR when the disease is rare.

Which measure of association is directly calculated from a cohort study, and why?

Relative risk (RR), because a cohort study follows exposed and unexposed groups forward and can measure the incidence (risk) of the outcome in each group. RR = incidence in exposed / incidence in unexposed.

What does a cross-sectional study measure, and what key limitation does it have for causal inference?

A cross-sectional study measures exposure and outcome SIMULTANEOUSLY at a single point in time (a 'snapshot'), so it estimates PREVALENCE and associations. Its key limitation: because exposure and disease are assessed at once, it usually cannot establish temporality (which came first), so it cannot establish causation.