Evidence Appraisal Hub

Navigator for the evidence-appraisal cluster — the machinery for moving from a body of studies to a graded recommendation. Nucleus: Certainty of Evidence vs Strength of Recommendation.

Framing the question

Framing a Decision Question — the four mandatory slots (PICO); why the comparator is the slot that most changes the answer; why study design is not pre-specified; how broadly to define the population before pooling becomes misleading
Rating Outcome Importance — critical / important / limited; only critical outcomes drive the verdict; choose outcomes by importance, never by what was measured; whose perspective decides

Rating Certainty of Evidence — the four grades; outcome-centric rating; design as starting point not verdict; the five downgrade and three upgrade factors; why it is explicitly not a quantitative system
Indirectness of Evidence — the four sources of indirectness; the gate between “it worked there” and “it will work here”, priced as lost certainty
Upgrading Observational Evidence — the three narrow exits from the observational floor: large magnitude, dose-response gradient, residual confounding pointing the wrong way
Surrogate Outcomes — when a marker may stand in for an outcome, and what it costs

Certainty of Evidence vs Strength of Recommendation — nucleus: certainty is one of four inputs to strength, and the off-diagonals are asymmetric; what sets strength once certainty is fixed; strength is not priority; recommendations are not dictates
Baseline Risk and the Relative-Absolute Split — why one relative effect can serve everyone while the recommendation still differs by group; the arithmetic-vs-effect-modification distinction

Saturated Fat Intake and Replacement — WHO 2023, the dietary-fat nucleus. Read here it yields three method findings as well as its recommendations: a conditional recommendation driven by certainty alone, a surrogate discounted at the strength step rather than the certainty step, and internally inconsistent certainty ratings under the guideline’s strongest replacement claim. RETRACTED at gate 1 (2026-07-26) — the ratings are per-design (Reynolds observational vs Hooper RCT), which is correct GRADE practice. See the page’s [REVIEW] marker.

Was GRADE Actually Used — the seven suggested criteria for claiming GRADE; how to turn a suspicion of under-argued method into a specific, citable finding
Which Objective Moved This Recommendation — a recommendation is a composite of evidence and non-evidence objectives; the evidence-to-decision table is where a body says which one moved it
Is the Food Category Doing Any Work — when the exposure is a category, does the boundary carry information? Sometimes yes (SSB inside sugars), sometimes demonstrably not (refined vs whole grain)
Physical Activity Dose and Mortality — where the PA mortality curve bends: ~24 min/day MVPA reaches maximal risk reduction (HR 0.34, HIGH certainty, accelerometer-measured); strength adds independently but is never ranked above aerobic; sedentary time carries its own thresholds
Whole Grains Refined Grains and Pulses — refined grains show a flat null (RR 1.00) for CVD and T2D; whole-grain benefit is cohort-only and SACN attributes it to cereal fibre; pulses largely unstudied and the nulls are low-power