Supplementary Table 2. The explanation for the factors given for the modified GRADE (Grading of Recommendations Assessment, Development and Evaluation) assessment.

	Downgrade factor					Upgrading factor
	Study limitation	Inconsistency	Imprecision	Indirectness	Reporting bias	Effect size	Dose-response
Description	Evidence base contains < 30% of all participants from studies rated overall low risk of bias (low quality in seven or more of the ten domains).	Pooled estimate has substantial or considerable heterogeneity (I2 > 50%), OR there is large variability in point estimates across studies, with estimates varying in direction and little or no overlap of confidence intervals.	95%CI are excessively wide about the pooled effect estimate, OR evidence base contains few studies with small populations and most provide imprecise estimates.	Differences in population, exposure, or outcome thought to substantially alter pooled estimate, limiting applicability and generalizability.	Evidence of reporting deficiencies and publication bias from funnel plot and Egger's test if ≥ 10 studies, OR the relationship has not been repeatedly investigated (< 10 studies).	Moderate to large pooled effect estimate (OR > 2.5).	Evidence of a well-defined pattern (linear or otherwise) suggesting an exposure response gradient is present.
Exception	Unless sensitivity analysis restricted to studies rated overall low risk of bias explains methodological heterogeneity. Then, report pooled estimate from overall low risk of bias studies.	Unless subgroup analyses explain clinical heterogeneity in pooled estimate. Then, report estimates from each subgroup.
Levels of confidence in the certainty of evidence
High	We have high confidence in the estimate: the true measure of association likely lies close to that of the estimate.
Moderate	We have moderate confidence in the estimate: the true measure of association is likely close to the estimate, but possibly substantially different.
Low	We have low confidence in the estimate: the true measure of association may differ substantially from the estimate.
Very low	We have very low confidence in the estimate: the true measure of association will likely differ substantially from the estimate.
Inconsistent	We could not conclude a consistent result based on all the investigations.
Limited	Only one study was included in the outcome.