Literature Review and Evidence Appraisal
Searching and selecting
Know how to frame a searchable question using population, intervention or exposure, comparator, and outcome. Understand the value of pre-registering a review protocol, using multiple databases, and documenting search strings and inclusion decisions so the process is reproducible. Distinguish a narrative review, a systematic review, and a meta-analysis: the first is selective, the second is systematic and transparent, and the third adds statistical pooling.
Appraisal concepts
Candidates must understand risk-of-bias domains for randomized trials (randomization process, deviations, missing data, outcome measurement, selective reporting) and for observational studies (confounding, selection, measurement). Learn to read effect estimates with confidence intervals and to interpret heterogeneity in a meta-analysis. Understand that a funnel plot asymmetry may reflect publication bias or other small-study effects, not bias alone.
Levels of evidence and grading
Know that study design contributes to but does not determine certainty; grading systems weigh risk of bias, inconsistency, indirectness, imprecision, and publication bias. Understand that a single large trial can outweigh several small ones, and that absence of evidence is not evidence of absence.
Traps
- Equating statistical significance with clinical importance.
- Assuming a meta-analysis is automatically higher quality than a large trial.
- Ignoring that a wide confidence interval crossing the null indicates imprecision, not proof of no effect.
- Treating a systematic review's conclusions as independent of its included-study quality.
Worked rule of thumb
Appraise in layers: is the question clear, was the search reproducible, were included studies at low risk of bias, and do the pooled estimates support the stated conclusion? If any layer fails, downgrade confidence in the finding accordingly.