Table 4.

Summary of Findings Table using the GRADE Rating System

Certainty assessment									Summary of findings
Outcome	Number of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence	Mindfulness intervention n	Control n	Relative effect (95% CI)	Absolute effect (95% CI)
Mindfulness training program vs. active control
Self-report anxiety	2	RCT	Seriousa	Very seriousb	Seriousc	Seriousd	Not suspectede	⨁◯◯◯ VERY LOW	109	102	-	SMDf 0.31 lower (0.90 lower to 0.28 higher)
Clinician-rated anxiety	1	RCT	Seriousg	N/Ah	Seriousc	Not seriousi	N/Aj	⨁⨁◯◯LOW	48	41	-	Gk 0.28 lower (0.13 lower to 0.70 higher)
Self-report depression	1	RCT	Seriousl	N/Ah	Seriousc	Not seriousi	N/Aj	⨁⨁◯◯LOW	61	61	-	Gk 0.45 higher (0.10 higher to 0.81 higher)
Self-report worry	2	RCT	Seriousm	Not serious	Seriousc	Seriousd	Not suspectede	⨁◯◯◯VERY LOW	80	80	-	SMDf 0.07 higher (0.26 lower to 0.40 higher)
Self-report trait mindfulness	2	RCT	Seriousn	Very seriousb	Seriousc	Seriousd	Not suspectede	⨁◯◯◯ VERY LOW	80	80	-	SMDf 0.16 higher (0.56 lower to 0.87 higher)
Self-report decentering	1	RCT	Not seriouso	N/Ah	Seriousc	Not seriousi	N/Aj	⨁⨁⨁◯ MODERATE	19	19	-	gk 1.31 higher (0.62 higher to 2.00 higher)
Blood markers of acute stress	1	RCT	Seriousp	N/Ah	Seriousc	Not seriousq	N/Aj	⨁⨁◯◯ LOW	42	28	-	Not estimable. Significant mindfulness training gains found for three measures
Amygdala fMRI	1	RCT	Seriousp	N/Ah	Seriousc	Seriousr	N/Aj	⨁◯◯◯ VERY LOW	15	11	-	Not estimable. Significant mindfulness training gains found for amygdala connectivity with frontal regions
Frontal cortex fMRI	1	RCT	Seriousp	N/Ah	Seriousc	Seriousr	N/Aj	⨁◯◯◯ VERY LOW	15	11	-	Not estimable. Significant mindfulness training gains found for activation in frontal regions
Mindfulness training program vs. inactive or non-specified control
Self-report anxiety	3	RCT	Very seriouss	Very seriousb	Seriousc	Not seriousi	Not suspectede	⨁◯◯◯ VERY LOW	109	103	-	SMDf 1.92 lower (3.44 lower to 0.40 lower)
Self-report depression	3	RCT	Very seriouss	Very serioust	Seriousc	Seriousd	Not suspectede	⨁◯◯◯ VERY LOW	109	103	-	SMDf 1.19 lower (3.11 lower to 0.73 higher)
Self-report worry	2	RCT	Very seriouss	Very seriousb	Seriousc	Seriousd	Not suspectede	⨁◯◯◯ VERY LOW	77	71	-	SMDf 2.27 lower (6.32 lower to 1.79 higher)
Self-report trait mindfulness	2	RCT	Seriousu	Very seriousb	Seriousc	Not seriousi	Not suspectede	⨁◯◯◯ VERY LOW	93	88	-	SMDf 0.85 higher (0.04 lower to 1.74 higher)
Default Mode Network fMRI	1	Non-randomised within-subjects waitlist control	Not seriousv	N/Ah	Seriousc	Not seriousw	N/Aj	⨁⨁⨁◯ MODERATE	32	32	-	Not estimable. Significant mindfulness training gains found for connectivity within DMN regions

^aEvidence was downgraded by 1 level because the overall risk of bias was rated as high in both studies, mainly due to missing outcome data, and there were some concerns in both studies

^bEvidence was downgraded by 2 levels as p < .050 and I² > 75%, indicating considerable heterogeneity

^cEvidence was downgraded by 1 level as there was not the availability of samples defined by sub-clinical symptoms of anxiety (e.g., generalised anxiety symptoms, trait anxiety) or trials that used a non-manualised or ‘stand-alone’ mindfulness intervention

^dDowngraded one level as pooled sample size meets power estimate but CI found to be wide (i.e., upper or lower limit cross SMD of 0.5 in both directions)

^eEgger’s test could not be conducted but publication bias was not strongly suspected as both negative and positive trial publications were found for outcomes, and a comprehensive search for studies was employed

^fSMD is the pooled estimate derived from the meta-analysis

^gEvidence was downgraded by 1 level because the overall risk of bias was rated as high due to missing outcome data

^hOnly one study is available for this outcome and therefore inconsistencies cannot be considered

ⁱPooled sample size meets power estimate and CI not found to be wide (i.e., upper or lower limit do not cross SMD of 0.5 in either direction)

^jCannot be assessed due to availability of only 1 study

^kg is derived from calculations using data available in the study article

^lEvidence was downgraded by 1 level, mainly due to missing outcome data

^mEvidence was downgraded by 1 level because the overall risk of bias was rated as high in one study, mainly due to missing outcome data, and there were some concerns in both studies

ⁿEvidence was downgraded by 1 level because the overall risk of bias was rated as high in one study, mainly due to missing outcome data, and there were some concerns in both studies

^oEvidence was not downgraded as risk of bias due to the lack of a pre-analysis plan was deemed unlikely to lower confidence in the estimate of effect for the outcome of interest

^pEvidence was downgraded by 1 level, mainly due to missing outcome data

^qSample size meets power estimate. Numerical value for CI not available but graph in journal article shows CI is not wide (do not cross mean difference of 0.5 in either direction)

^rCI not available but sample size does not meet power estimate

^sEvidence was downgraded by 2 levels because the overall risk of bias was rated as high in two studies, mainly due to missing outcome data, and there were some concerns in all studies. One study had a high risk of bias or some concerns in all domains

^tEvidence was downgraded by 2 levels as p < .050 and I² > 75%, indicating considerable heterogeneity. When an outlier study that did not specify the control that was used was removed from the meta-analysis, heterogeneity was no longer observed

^uEvidence was downgraded by 1 level, mainly due to missing outcome data

^vEvidence was not downgraded as risk of bias due to confounding or the lack of a pre-analysis plan was deemed unlikely to lower confidence in the estimate of effect for the outcome of interest

^wCI not available but sample size meets power estimate