Table 3.
Random-effects meta-analysis results of randomized and non-randomized controlled trials
| Meta-analysis sub-groupa | No. participants/trials/studies | No. (%) trials with combined outcome measuresb | Effect measure | Effect estimate average (95 % CI)M/median (range)R,c | I 2 | GRADEd |
|---|---|---|---|---|---|---|
| RCT | ||||||
| Educational training/courses | ||||||
| Adults-Behaviourse | 709/4/4 | 2 (50) | SMD | 0.68 (−0.06, 1.41)R | 94 % | Low |
| Adults-Knowledge | 596/3/3 | 0 (0) | SMD | 0.87 (−0.05, 1.29)R | 93 % | Very low |
| Children/youth-Behaviours | 6379/2/2 | 1 (50) | SMD | 0.20 (0.05, 0.35)R | 96 % | High |
| Media campaigns/other messaging | ||||||
| Adults-Behavioural intentions | 117/2/2 | 1 (50) | SMD | 0.36 (0.02, 0.69)M | 0 % | Moderate |
| Adults-Behaviours | 686/4/4 | 1 (25) | SMD | 0.24 (−0.17, 1.03)R | 85 % | Low |
| Adults-Knowledge | 528/3/3 | 0 (0) | SMD | 0.42 (0.03, 0.92)R | 82 % | Low |
| Adults-Attitudes | 4914/8/8 | 4 (50) | SMD | 0.34 (0.05, 0.76)R | 94 % | Low |
| NRT | ||||||
| Educational training/courses | ||||||
| Adults-Behaviourse | 1099/4/2 | 0 (0) | SMD | 0.37 (0.08, 0.66)M | 58 % | Low |
| Adults-Knowledge | 1356/5/3 | 0 (0) | SMD | 0.44 (0.12, 1.14)R | 82 % | Low |
| Adults-Attitudes | 778/4/2 | 1 (25) | SMD | 0.26 (0.10, 0.43)M | 0 % | Moderate |
| Children/youth-Behaviours | 329/3/2 | 0 (0) | SMD | 0.33 (0.17, 0.90)R | 64 % | Very low |
| Children/youth-Knowledge | 339/3/2 | 0 (0) | SMD | 0.24 (0.14, 0.73)R | 75 % | Very low |
| Media campaigns/other messaging | ||||||
| Adults-Behaviours | 1118/2/2 | 0 (0) | RR | 2.31 (1.30, 3.33)R | 90 % | Low |
| Adults-Attitudes | 1442/3/3 | 1 (33) | RR | 1.75 (1.01, 2.85)R | 95 % | Very low |
RCT randomized controlled trials, NRT non–randomized controlled trials, SMD standardized mean difference (Hedges’ g), RR relative risk, CI confidence interval
aSubgroups divided by study design, intervention type, target population, and outcome type
bRefers to studies that reported multiple measures of the same construct in the same individuals, which were combined post hoc into one overall measure
cSuperscript M indicates that an average estimate of effect and 95 % CI is provided because heterogeneity was low to moderate (I 2 = 0–60 %). Superscript R indicates that the median and range of study effect sizes is provided because heterogeneity was high (I 2 > 60 %)
dExplanation of the GRADE ratings:
Very low = the true effect is likely to be substantially different from the measured estimate
Low = the true effect may be substantially different from the measured estimate
Moderate = the true effect is likely to be close to the measured estimate, but there is a possibility that it is substantially different
High = strong confidence that the true effect lies close to that of the measured estimate
eOne trial/study in each of these analyses used an imputed pre-post correlation value of 0.83 from Kendall et al. (2004) [37]. In both cases, sensitivity analyses indicated that the selection of the imputed value had an appreciable impact on the meta-analysis results (Additional file 12), leading to a downgrading of these findings in the GRADE assessment (Additional file 10)