Table 1.

Criteria for judging risk of bias of selected studies, adapted from the criteria suggested by the Cochrane Handbook for systematic reviews of interventions

Selection bias of participants
‘Low risk’ of bias	Sites Multicentred Bicentred
	Recruitment Random selection Sequential selection (e.g. first 50 volunteers) Patient list from set criteria (e.g. age)
	Inclusion/Exclusion Children of parents in the study group have cancer diagnoses Homogeneous times since diagnoses
‘High risk’ of bias	Sites Monocentred
	Recruitment Convenience sample
	Inclusion/Exclusion Children of parents in the study group have heterogeneous diagnoses (cancer and other condition) Heterogeneous times since diagnoses
‘Unclear risk’ of bias	Insufficient information to permit judgement of ‘Low risk’ or ‘High risk’
Attribution bias (incomplete outcome data from participants)
‘Low risk’ of bias	Statistical comparisons between participants and non‐participants High response rate (questionnaires: ≥65%; interviews: ≥75%) Low attrition (according to number of measures used and length of follow‐up) Missing data identified and managed for analysis
‘High risk’ of bias	Low response rate (questionnaires: 65%; interviews: 75%; Arber, 2001) High attrition (according to number of measures used and length of follow‐up) Missing data identified but not managed for analysis
‘Unclear risk’ of bias	Insufficient information to permit judgement of ‘Low risk’ or ‘High risk’
Reporting bias (from authors) due to selective outcome reporting
‘Low risk’ of bias	Objectives/Hypotheses Objectives/hypotheses clearly stated Objectives/hypotheses addressed accordingly
	Demographics Demographic information on participants available
	Outcome data/confounds All collected data available in the results section All data acknowledged (whether approving or disproving the hypotheses) Confound variables acknowledged and controlled for statistically (e.g. as covariate in multivariate analyses) Reasons for missing data acknowledged Use of multi‐informant strategy when collecting data in children Consideration of effect sizes (e.g. Cohen's d)
	Discussion/Conclusions Conclusions based on results Limitations acknowledged If preliminary study/pilot: acknowledged No conflict of interest (source of funding acknowledged)
‘High risk’ of bias	Objectives/Hypotheses Objectives/hypotheses not clearly stated Not all of the study's objectives/hypotheses addressed
	Demographics Missing demographic information on participants
	Outcome data/Confounds Missing collected data in the results section Missing data disproving the hypotheses Confound variables not acknowledged No acknowledgement of reasons for missing data No use of multi‐informant strategy when collecting data in children No consideration of effect sizes (Cohen's d)
	Discussion/Conclusions Conclusions depart from results Limitations are not acknowledged If preliminary study/pilot: not acknowledged Conflict of interest (source of funding not acknowledged)
‘Unclear risk’ of bias	Insufficient information to permit judgment of ‘Low risk’ or ‘High risk’ Confound variables acknowledged but not controlled for statistically
Other bias
‘Low risk’ of bias	Participants Large sample (Ratio of participants to predictors satisfying Tabachnick & Fidell's (2007) guidelines: N > 50 + 8p, where p: number of predictors)
	Measures Structured clinical interview Measures with satisfactory pre‐established psychometric properties Multimodal assessment (e.g. medical records and self‐reports)
	Procedures Use of procedures to maximise response rate or limit attrition
	Design Longitudinal Data analysis strategy appropriate for exploring longitudinal causal effects (e.g. predicted outcome controlled for at baseline)
‘High risk’ of bias	Participants Small sample (Ratio of participants to predictors not satisfying Tabachnick & Fidell's (2007) guidelines: N > 50 + 8p, where p: number of predictors)
	Measures Measures have not been validated or insufficient pre‐established psychometric properties (e.g. ad hoc or home‐made measures) Composite measures (e.g. combining subscales of parental distress and other constructs)
	Procedures No procedures used to limit attrition
	Design Cross‐sectional
‘Unclear risk’ of bias	Insufficient information to permit judgement of ‘Low risk’ or ‘High risk’

Arber S. (2001) Designing samples. In: Researching Social Life (ed. Gilbert N). SAGE Publications, London.

Tabachnick B.G. & Fidell, L.S. (2007) Using Multivariate Statistics, 5th edn. Pearson, Boston, MA.