Table 5

Risk of bias in randomised controlled trials

Selection biasPerformance biasDetection biasAttrition biasSelective outcome reportingOther bias
StudyRandom sequence generationAllocation concealmentBlinding of participants + personnel (subjective/objective outcomes)Blinding of outcome assessment (subjective/objective outcomes)Incomplete outcome data (subjective/objective outcomes)
Beresford et al 46 No detailsNo detailsBlinding not possibleNo details
Lack of blinding may affect subjective outcomes
Numbers provided but reasons not fully explainedOutcomes reported as in method
No protocol
No other bias
Racette et al 35 No detailsNo detailsBlinding not possibleNo details
Lack of blinding may affect subjective outcomes
Numbers and reasons providedNo protocolRandomisation was only over two sites
Demographical difference between groups
Sorensen et al 47 No detailsNo detailsBlinding not possibleNo detailsNumbers provided and gives some reason for questionnaires being eliminated from analysesFood Frequency Questionnaire was used but only fibre and fat reportedNo other bias
Sorensen 36 No detailsNo detailsBlinding not possibleNo detailsNot reported but analysis presented for all sample (not restricted for completers)Additional dietary data, not presented here, were collected by means of the Food Frequency QuestionnaireNo significant differences among the groups at baseline
All analyses were computed by taking into consideration the nesting of employees in worksites
Stites 37 Using a computer programNo detailsBlinding not possibleNo detailsMainly reported and reasons explained, for one participant, no reason explainedOutcomes reported as in method
No protocol
No other bias
Vermeer 42 No detailsNo detailsBlinding not possibleNo detailsNumbers provided but reasons not fully explainedOutcomes reported as in method
No protocol
No other bias
  • Green, low risk of bias; orange, unclear risk of bias; red, high risk of bias; ‘quote from publication’.