This page explains the research vocabulary that appears throughout evidage — terms like meta-analysis, RCT, hazard ratio, and so on — without assuming a science background. Think of it as a reading companion you can come back to whenever you hit unfamiliar terminology.
📚 The Hierarchy of Evidence
Not all studies are equal. The general hierarchy from strongest to weakest:
- Meta-analyses & Systematic Reviews (Level 1): Pool results from many studies for a more reliable answer.
- Randomized Controlled Trials (RCTs) (Level 1–2): The gold standard for testing cause and effect.
- Cohort Studies (Level 2): Follow large groups over time. Suggest associations, not proof of cause.
- Case-Control Studies (Level 3): Compare people with and without a condition retrospectively.
- Animal & Cell Studies (Level 4): Useful for mechanism, but rarely reliable for human behavior change.
- Expert Opinion (Level 4): The weakest form, though still informative when stronger evidence is missing.
🔍 Key Terms Explained
Meta-Analysis
A statistical pooling of multiple individual studies to produce a single, more precise estimate. If 20 RCTs each tested whether avocados lower LDL, a meta-analysis combines them all and tells you “yes, by about 4 mg/dL on average.” Meta-analyses are powerful because random errors in individual studies cancel out.
Randomized Controlled Trial (RCT)
Participants are randomly assigned to receive either the intervention (e.g., one avocado per day) or a control (e.g., usual diet). Random assignment minimizes bias. RCTs are the strongest design for proving that A causes B, not just that A and B happen together.
Cohort Study
Researchers follow a large group of people over years or decades and compare outcomes between subgroups. Famous examples include the Nurses’ Health Study and the Japan Public Health Center-based Prospective Study (JPHC). Cohort studies are great for revealing patterns over time but cannot prove causation by themselves.
Hazard Ratio (HR)
How much more (or less) likely an event is to occur in one group compared to another. HR=1.0 means no difference. HR=0.7 means 30% lower risk. HR=1.5 means 50% higher risk. Always check the confidence interval — if it crosses 1.0, the result may not be statistically significant.
Relative Risk vs. Absolute Risk
“Reduces risk by 50%” sounds dramatic, but it depends on the baseline. If your baseline 10-year risk is 2%, a 50% reduction takes you to 1% — meaningful but small. Always look at absolute risk reduction, not just relative.
P-value
The probability that the observed result happened by chance, assuming no real effect exists. p<0.05 is the conventional threshold for "statistically significant" — meaning a less than 5% chance of being a fluke. P-values say nothing about how large or important an effect is.
Confidence Interval (CI)
A range that captures where the true value likely lies. “HR 0.85, 95% CI 0.75–0.95” means we’re 95% confident the true hazard ratio is between 0.75 and 0.95. Narrow intervals indicate precise estimates; wide intervals indicate uncertainty.
GRADE Framework
An internationally adopted system for rating the certainty of medical evidence. evidage’s Level 1–4 labels map roughly to GRADE’s High/Moderate/Low/Very Low certainty grades.
🚨 Red Flags in Health Reporting
When you see these patterns, slow down before believing the headline:
- “Studies show…” with no citation. Always check what study, what design, what sample size.
- Single small study being presented as final truth. Replication matters.
- Animal study framed as a human breakthrough. Most don’t translate.
- Sponsored research from companies that profit from the result. Funding source matters.
- Surrogate endpoints (e.g., “lowers cholesterol”) rather than real outcomes (e.g., “reduces heart attacks”). Surrogates are easier to measure but don’t always translate.
evidage applies these checks for you, every time.