Evidence-Based Practice, Literature Evaluation, and Research Application
What this domain tests
Literature evaluation items ask you to decide not just whether a study is statistically significant, but whether it is believable, applicable, and clinically meaningful. Expect questions on trial design, bias, endpoint selection, and - most heavily - number needed to treat, absolute risk, and confidence interval interpretation.
Concepts to master
Use PICO to frame a question and the evidence hierarchy to rank answers (systematic review/meta-analysis of RCTs > RCT > cohort > case-control > cross-sectional > case series > expert opinion). Assess internal validity through randomization, allocation concealment, blinding, completeness of follow-up, and intention-to-treat analysis. Distinguish surrogate endpoints (LDL, A1c, tumor response) from patient-oriented outcomes (mortality, MI, stroke, hospitalization, quality of life). Understand noninferiority designs - they require a pre-specified margin and ideally both ITT and per-protocol analyses - and know why a composite endpoint can be driven entirely by its softest component. Recognize heterogeneity (I²) and funnel-plot asymmetry as signals about pooling and publication bias.
Common traps
Relative risk reduction sounds large and is often clinically small. Investigators may foreground per-protocol results, post-hoc subgroups, or unadjusted analyses. Odds ratios are not risk ratios and overstate effects when events are common. A p-value of 0.04 in a trial with a dozen endpoints is likely multiplicity, not discovery. Read the confidence interval width, not merely whether it crosses 1.
Worked example / rule of thumb
Control event rate 20%, treatment event rate 15%. RRR = (20 − 15)/20 = 25%; ARR = 5%; NNT = 1/0.05 = 20. Rule of thumb: convert every relative measure into an absolute one before judging value, and always ask "NNT over what time horizon, in which patients?"