Research and Evidence-Based Practice
Appraising evidence
Know the hierarchy: systematic reviews and meta-analyses of RCTs at the top, then RCTs, cohort studies, case-control, cross-sectional, ecological, and case reports or mechanistic reasoning at the bottom. A meta-analysis is only as trustworthy as the studies it pools - heterogeneity, publication bias, and risk of bias must be assessed.
Threats to validity: selection bias, measurement and recall bias, attrition, and confounding (internal validity) versus generalizability (external validity). Association is not causation; consider temporality, dose-response, plausibility, and consistency.
Statistics: a p-value is the probability of the data given the null, not the probability the null is true, and it says nothing about magnitude. Report effect sizes with confidence intervals (RR, OR, HR, mean difference) and absolute measures such as NNT. Screening metrics: sensitivity rules out (a negative result from a highly sensitive test is reassuring), specificity rules in; positive predictive value rises as prevalence rises.
DRI derivation: EAR from balance and status data; RDA = EAR + 2 SD; AI when an EAR cannot be determined; UL from no-observed-adverse-effect levels with uncertainty factors; AMDR as a percentage of energy.
Worked example: a test with 90% sensitivity and 90% specificity in a population with 10% prevalence. Of 1,000 people: 90 true positives and 90 false positives → PPV = 90 ÷ 180 = 50%. Even a good test yields mostly false positives when prevalence is low.
Trap: choosing an "authoritative" systematic review of observational studies over a well-conducted RCT when the question is about treatment effect.