Number of Studies	Study Design	Quality of Studies	Consistency	Directness	Other Modifying Factors
N	RCT=High Observational = Low Any other evidence = Very Low	Serious limitation to study quality (−1) Very serious limitation to study quality (−2)	Important inconsistency (−1)	Some uncertainty about directness (−1) Major uncertainty about directness (−2)	Association strong (+1) Association very strong (+2) Dose response gradient (+1) All plausible confounders would have reduced the effect (+1) Imprecise or sparse data (−1) High probability of reporting bias (−1)

Source: Atkins D et al. Grading quality of evidence and strength of recommendations. BMJ 2004;328(7454):1490. (46)