Research Methodology
Research methodology tests whether you can judge study design and avoid overreading results. The exam expects clear questions, appropriate comparators and recognition of bias.
Core rules
- Randomized trials best assess causal treatment effects; observational studies assess association and need confounding control.
- Bias includes selection, information, confounding and reporting bias; matching and multivariable adjustment reduce but do not eliminate confounding.
- Endpoints must be clinically meaningful; surrogate endpoints may not predict patient-important outcomes.
- Sample size and power depend on effect size, variability and alpha; underpowered negative studies are not proof of no effect.
Worked example
A single-center review reports fewer infections with a new protocol, but compliance was higher in the protocol group. The reasoning should identify confounding and implementation bias; a randomized or interrupted time-series design would better test causality.
Traps
Do not call association causation; do not accept surrogate endpoints as definitive; do not ignore loss to follow-up; do not interpret a nonsignificant underpowered trial as equivalence.