Table 3.

Intervention Participation, Adherence, Effects & Limitations in Study Quality

				Limitations in Study Quality
Author et al., Year	Participation & Adherence	Impact on Sleep Outcomes (Subjective, Objective, Participant Ratings)	Impact on Other Outcomes	Study Design	Outcome Assessment & Follow-Up	Quality of Intervention Implementation	Participants & External Validity
Antonson et al., 2018	Intervention Participation: Of the 95 randomized to iMBI, only 14 completed ≥ 1 session, of whom 4 completed ≥10 sessions. Only 1 person completed the iMBI intervention, defined as completing ≥40 of 96 planned sessions. Research Protocol: Only 3 of the 4 the participants who completed ≥10 iMBI sessions completed post-intervention assessments. No association between compliance and stress or psychiatric symptoms.	Insufficient statistical power to examine sleep outcomes. The only participant who completed iMBI sessions decreased PSQI global score 8 to 7.	Only participant who completed iMBI sessions decreased GSI by 31% (from 0.80 to 0.55), PSS from 29 to 23.		Insufficient statistical power given low adherence. No follow-up assessment of outcomes.	Low intervention participation and adherence.	Low uptake of the intervention at the recruitment school. Consented only 283 out of 1403 (20%) and only 202 completed baseline surveys (71% of those consented or 14% of the school sample).
Bartel et al., 2018	Intervention Participation: Self-reported. Classified as completer if 3 or more nights intervention for 1 or more weeks 71% (50/70) completed Mindfulness.	No significant time arm*time interaction for SOL, overall (p=0.17) or for the subset (n=119) who had SOL of 30 minutes or more at baseline (p=0.07). Mindfulness arm reduced SOL only among those with long SOLs, if performed for at least 3 nights/week. For the group of completers with baseline SOL ≥ 30 minutes, significant arm*time interaction (p=0.001): 65.35 to 34.60 minutes. No significant findings for sleep hygiene (ASHS). Perceived Effectiveness: Perceived efficacy of intervention among mindfulness participants: problems falling asleep improved a lot (31.6%) or a bit (45.5%). Intervention was highly (6.1%) or somewhat effective (38.8%).	No significant impacts on rumination (PTQ-C).		Did not objectively measure sleep. Social desirability in reporting possible. Assessment at mid-intervention but no follow-up assessment of outcomes.	Adherence (subjectively measured) was low. Mindfulness audio needed “improving” according to adolescent participant qualitative feedback.	Listening to audio not mindfulness per se may be operating.
Bei et al., 2013	Intervention Participation: 90% completion rate (completed 5 of 6 planned sessions). Research Protocol: One actigraph did not work; 8 participants had actigraphy data.	Objective: Moderate improvements for SOL (Cohen’s d = 0.53; 17 min decrease) and SE (Cohen’s d = −0.51). Participants who took ≥30 minutes to fall asleep fell from 70% to 35.75% post-intervention. Small improvement for total sleep time (Cohen’s d = −0.46, 20 more minutes of sleep/night). Changes in regularity of bed time and rise time were small (Cohen’s d = 0.27 & −0.21, respectively). Post-intervention, participants went to bed on average 35 minutes earlier and woke up 40 minutes earlier. No significant impact on WASO. Subjective: PSQI global scores improved moderately (Cohen’s d = 0.51). Large improvement in subjective sleep quality (Cohen’s d = 0.84). SOL had large improvement (Cohen’s d = 0.80; 19 minute reduction). Participants who took ≥30 minutes to fall asleep reported taking 30 min or longer to fall asleep decreased from 90% to 55.56%. No significant difference in self-reported: TST, SE, subjective sleep disturbance, or sleep-related daytime dysfunction. Perceived Effectiveness: Out of 7, the average rating of overall perceived helpfulness was 5.44 (SD = 0.73).	No statistically significant differences in anxiety or mood disorders.	Single-arm design without control group.	Adherence was not measured. No follow-up assessment of outcomes. Pre- and post-intervention assessments were both measured during holidays.		Small sample size. Only females. Not clear what most effective component of intervention was (it was multi-component).
Blake et al., 2016	Intervention Participation: Participants attended approximately 76% of sessions on average. In total 5 participants (4 intervention, 1 control) did not complete at least 4 of the 7 intervention sessions. Research Protocol: 20 participants (10 intervention, 10 control) were randomized but did not initiate intervention. Low missing data for questionnaire (6.2%) and actigraphy (6.1%), slightly higher for sleep diary variables (14.6%). Participants wore the actiwatch on 4.5 of the 5 nights. Participants completed sleep diaries on 3.75 of the 5 school nights. Of participants, 4 declined to provide post-intervention data.	Objective: Significantly improvements compared to control in objective SOL. No improvements in WASO. Subjective: Significant improvements in subjective SOL, subjective sleep quality (PSQI global score), subjective daytime sleepiness. No significantly greater improvements in objective or subjective total sleep time. Perceived Effectiveness: Believed got something of value: Sleep SENSE 100% vs. Study SENSE (active control) 96%. Sleep SENSE participants rated the program as more useful than controls (p<.01). Sleep participants reported practicing mindfulness (5 or more minutes a time) 1 – 2 times/week. Most helpful sleep components: mindfulness of the breath, going to bed.	After controlling for baseline values, compared to controls, immediate post-intervention, sleep participants had less anxiety, but not less depression.				A multi-component intervention: canť tease out which components worked. Those with most severe symptoms were more likely to enroll and be engaged once enrolled. Excluded those with major depressive disorder.
Blake et al., 2017	As above (Blake 2016).	Objective: Assessed direct effects using modified intent to treat (including non-completers but not non-initiators). Intervention participants had shorter sleep onset latency measured with actigraphy (p=0.004). Subjective: Better sleep efficiency measured with diary (b=9.97, p=0.044), and better global sleep quality (PSQI b=1.07, p=0.001) compared to control participants. Mediation: Assessed mediation - found significant indirect effects for sleep quality (PSQI) and anxiety (SCAS) through improvements in presleep arousal (PSAS), somatic and cognitive subscales.	Intervention resulted in: less anxiety symptoms, less pre-sleep somatic arousal, less presleep cognitive arousal and better sleep hygiene awareness.				No dose-response effect observed. Did not detect specific effects based on intervention components.
Blake et al., 2018	As above (Blake 2016).	Moderation: Assessed gender and baseline sleep duration, anxiety, depression, and self-efficacy as moderators. For PSQI, anxiety, depression, and self-efficacy were moderators (stronger intervention effects among those with higher anxiety and higher depression and higher self efficacy scores at baseline), but not gender. No significant moderation on sleep onset latency (SOL) or sleep efficiency (SE) for any moderators tested.	Low levels of self-efficacy may need additional targeted support				Common variance across the moderator variables. Other possible moderators (such as age) not assessed. Excluded those with major depressive disorder.
Hirokawa et al., 2012	3 participants (2 intervention, 1 control) dropped from intervention.	“Improved sleepiness at awakening” (mean score 12.2 pre, 16.3 during, 17.7 post for intervention; 13.7, 11.2, 12.1 respectively for control; p<0.01); “experiencing persistence of fatigue” (11.9, 16.9, 20.4 for intervention; 18.2, 12.4, 14.6 for control, p<0.001); “being able to formulate an answer” (12.5, 17.1, 15.1 for intervention, 10.3, 8.7, 9.4 for control, p<0.01). No significant group X time effects for alleviation of tiredness, longer duration of sleep, or sleep hours.	No significant effect of sex for quality of sleep on any subscores. No difference in effects across the 5 days of the intervention.	Used water as a comparator, rather than an aroma other than lavender.	No objective measurement of sleep outcomes. Social desirability response bias is possible. No follow-up assessment of outcomes. Sleep diary not analyzed due to privacy concerns. Did not assess known factors related to sleep quality (caffeine, being active before). Did not assess any outcomes other than sleep.	Intervention was delivered by study team interventioninst, but the mind-body training for the interventionist was not described. Adherence to intervention was not assessed.
Jeitler et al., 2017	Research Protocol: 5 participants (4 intervention, 1 control) lost to follow-up by immediate end of intervention; 33 (22 intervention, 11 control) lost to follow-up at 6 month intervention.	No significant improvements from baseline in sleep (as measured with visual acuity scale) immediately post-intervention or at 6-month follow up	No significant differences in PSS between groups immediately post-intervention or at 6-month follow up. Twice as many side effects reported in the yoga intervention group compared to control, with 4 reporting headache after yoga class.	Groups were not comparable due to lack of randomization. Intervention group had higher baseline scores of perceived stress, depression and anxiety. Intervention participants attended school sports (control condition) after the 10 week intervention period.			Substantial loss to follow up at 6 months.
Manzaneque et al., 2009	Intervention Participation: 6 participants (5 intervention, 1 control) did not complete intervention.	Subjective: No significant difference in post-intervention PSQI global score (4.84 intervention vs. 5.27 control, p=0.83), but statistically significant difference in sleep duration.	No significant differences in post-intervention serum cytokines (TNF-alpha or IFN-gamma). Significantly (p<0.01) better outcomes for intervention for anxiety (STAI) and depression (BDI).	Comparator was not active control (“follow daily life habits”). Small sample size.	No objective measurement of sleep outcomes. No follow-up assessment of outcomes.		Participants were general healthy population, not those with sleep problems at baseline.
Swanson et al., 2016	Intervention Protocol: Low adherence (3 youth completed one or two sessions and 8 youth completed three to four sessions) Research Protocol: In total, 12 of 22 had complete pre/post data; 8 youth had no post-surveys or session attendance data.	Subjective: Among those with complete data (n=12), overall sleep quality improved significantly (PSQI-Q6 change = 0.67;p< .05).	Non-significant improvement in mental distress (CES-D change = 4.16,p= 0.036) or self-compassion SCS-SF change = 0.11,p= 0.64). No significant change in mindful state (MAASA p=0.86).	Small sample size. Single-group design without comparator.	No objective measurement of sleep outcomes. No follow-up assessment of outcomes.	Low adherence to intervention and high attrition due to unstable housing.
Vera et al., 2019	Research Protocol: 5 lost to follow-up. No statistically significant differences in baseline char between those with and without complete data.	Subjective: No significant difference in subjective sleep quality (3.63 in Control vs. 3.48 in intervention - p=0.77). Notably, scores are lower (indicating better sleep) compared to baseline PSQI (4.65, 5.14 respectively).	In the intervention group, compared to controls: significantly lower ACTH (19.65 vs. 24.06, p=0.02), but no significant differences in Cortisol, DHEA-S, or measures of anxiety (STAI) or depression (BDI)	Small sample size. Comparator was not active control (“regular daily life habits”)	No objective measurement of sleep outcomes. No follow-up assessment of outcomes.	Short duration of intervention.
Zelic et al., 2018	Intervention Protocol: Experimental group had 82% compliance in listening to intervention. Research Protocol: 65 of 86 participants provided ≥3 days of data (in 7 days intervention) and were retained in analyses. Three participants withdrew from study	No significant between-group differences between the mindfulness and control conditions on any measure including objective or subjective sleep indices. Assessed depression, rumination, anxiety, worry, and gender as moderator of sleep outcomes. Participants with poor baseline overall sleep quality in the mindfulness condition had longer sleep duration (p <.05). High ruminators in the mindfulness condition obtained significantly more sleep time (p < .05). Females in the mindfulness condition obtained a significantly greater amount of sleep compared to females in the control condition (p < .05), but no significant effect in males.	No significant between-group differences between on non-sleep outcomes.	Comparator was not active control (“no change to sleep routine”)		Low adherence led to insufficient sample size. Short study duration/low intensity.	Homogeneous sample. Barriers to recruitment due to skepticism about possible Buddhism connotation of mindfulness.

Abbreviations: ASHS = Adolescent Sleep Hygiene Scale; BDI = Beck Depression Inventory; CES-D = Center for Epidemiological Studies – Depression; GSI = general psychiatric symptoms subscale of Symptoms Checklist-90; iMBI = Internet-based mindfulness-based intervention; MAASA = Mindful Attention Awareness Scale; PSAS = Pre-sleep Arousal Scale; PSQI = Pittsburgh Sleep Quality Index; PSS = Perceived Stress Scale; PTQ-C = Perseverative Thinking Questionnaire, children; SCAS = Spence Children’s Anxiety Scale; SCS-SF = Self-Compassion Scale, Short Form; SE = sleep efficiency; SOL = Sleep onset latency; STAI = State-Trait Anxiety Inventory; TST = total sleep time; WASO = wake after sleep onset