Study | Random sequence generation | Allocation concealment | Blinding of participants and personnel | Blinding of outcome assessment | Incomplete outcome data | Selective reporting | Other bias | Summary |
---|---|---|---|---|---|---|---|---|
Ajilchi et al. (2019) [45] | cRandom draw | aRandomisation conducted by an independent party unconnected to project | cStudy was non-blinded | cStudy was non-blinded | aEach participant completed the intervention | aAll prespecified outcomes were reported | bAuthors transparent throughout. Small sample that is not diverse | High risk of bias for this study. Three domains showed a high risk of bias due to high risk of selection, detection and performance bias |
Donohue et al. (2018) [47] | aUrn randomisation | bUnclear who performed randomisation | bNo measures described to blind participants to intervention | aAssessors from clinic that operated independently from intervention programmes. No blinds assessed to be broken. | aAnalyses adjusted for data being missing at random | aAll outcome measure effects were reported, along with effect sizes for each group | aStudy limitations addressed and transparency ensured throughout | Low risk of bias for this study. Two domains were unclear but they were unlikely to have had major bearing on results |
Glass et al. (2019) [51] | aStratified random sampling to ensure comparable groups | bUnclear who performed randomisation | bParticipants were asked not to discuss details of intervention but it is possible discussion took place as participants were students at same university | bUnclear whether assessors had knowledge of treatment groups when assessing effects | cHigh attrition rate leaves data susceptible to attrition bias | aAll outcome measure effects were reported, along with effect sizes for each group | aStudy limitations were highlighted | Unclear risk of bias for this study. Aspects of selection, performance and detection bias were unclear. High risk of attrition bias due to nature of mindfulness intervention |
Gross et al. (2018) [44] | cAttempt was made to use random selection but due to time constraints it was not employed, decision was taken to use one team. | bUnclear who performed randomisation into the two intervention groups | bParticipants were from the same team so there was potential for discussion about details of intervention | bUnclear whether assessors had knowledge of treatment groups when assessing effects | aAttrition and losses to follow-up were disclosed and reasons provided | aAll prespecified outcomes were reported | bPotential for allegiance effects influencing results as one of the groups was led by an author of the study but the study showed that therapeutic rapport did not have a significant effect | Unclear risk of bias for this study. Lack of randomisation raises prospect of selection bias but overall the process was transparent |
Gulliver et al. (2012) [63] | aAutomated computer system used | aConditions allocated by researchers not involved in day-to-day management | aDescribed method used to reduce likelihood of participant knowledge of intervention | bUnclear whether assessors had knowledge of treatment groups when assessing effects | aAnalyses adjusted for data being missing at random | aAll outcome measure effects were reported, along with effect sizes for each group | aStudy limitations were addressed and caution is urged when interpreting significant effects | Low risk of bias for this study. One domain (blinding of outcome assessors) was unclear but it is unlikely if that influenced the results given the online format of the intervention and data collection |
Liddle et al. (2019) [40] | aRandomisation occurred using a random number generator | aRandomisation conducted by an independent researcher not involved in intervention or data analysis | aParticipants not informed of allocated condition | bUnclear whether assessors had knowledge of treatment groups when assessing effects | aAnalyses adjusted for data being missing at random | aAll prespecified outcomes were reported | aAuthors were transparent about each stage of the intervention design | Low risk of bias for this study. One domain (blinding of outcome assessors) was unclear but it is unlikely to have significant impact on results |
Sekizaki et al. (2019) [55] | aRandomisation was performed using each student’s school number | bUnclear who performed randomisation | cStudy was non-blinded and in the same school there was risk for sharing of information between groups | cStudy non-blinded, potential for detection bias | aEach participant completed the intervention | aAll prespecified outcomes were reported | aStudy limitations were addressed and authors urged caution over the generalizability of the findings | Moderate risk of bias for this study. Selection, attrition and reporting bias risk was low. Risk of performance and detection bias was high due to no blinding. |
Van Raalte et al. (2015) [69] | bMethod not disclosed | bUnclear who performed randomisation | bUnclear if participants were or were not blinded to their intervention | bUnclear whether assessors had knowledge of treatment groups when assessing effects | aAnalyses adjusted for data being missing at random | aAll outcome measure effects were reported, along with effect sizes for each group | aAuthors were transparent about each stage of the intervention design | Unclear risk of bias for this study. Information on selection, performance and detection bias was not disclosed, though attrition and reporting bias was low |
Summary of bias across studies | Random sequence generation was performed in each study bar one. One study did not disclose method | Methods of allocation were largely unclear except for three studies where risk of bias was low | Blinding of participants was mixed, 4 studies were unclear while two had high risk and two low | The risk of bias was mixed, 5 studies were unclear while two had high risk and one low for blinding the assessors’ knowledge | 7 of 8 studies displayed low risk of bias for controlling missing data, one study was high | There was a low risk of bias across the studies for reporting outcomes | Transparency was ensured by each of the studies, resulting in a low risk of bias for 6 studies and 2 unclear | Risk of selection, performance and detection bias findings were mixed. The risk for attrition and reporting bias was low with transparency maintained throughout each of the studies |