Thresholds for statistical and clinical significance in systematic reviews with meta-analytic methods

Table 1 Quality assessment criteria according to GRADE adopted after reference[16]

Study design	Levels of confidence in estimate	Decrease confidence estimate if	Increase confidence estimate if
Randomised clinical trials	High	Risk of bias	Large intervention effect
Randomised clinical trials	High	One level if serious	One level if large
		Two levels if very serious	Two levels if very large
	Moderate	Imprecision	Dose response
		One level if serious	One level if evidence of dose response
		Two levels if very serious
Observational studies	Low	Indirectness	All plausible confounding
		One level if serious	One level if confounding would reduce a demonstrated effect
		Two levels if very serious
	Very low	Heterogeneity	All plausible confounding
	Very low	One level if serious	One level if confounding would suggest a spurious effect when results show no effect
		Two levels if very serious
		Publication bias
		One level if serious
		Two levels if very serious

ISSN: 1471-2288