Table 2.
Assessment of risk of bias of included studies
| Study | Random sequence (Judgment) | Allocation concealment (Judgment) | Blinding participants (Judgment) | Blinding outcome (Judgment) | Incomplete outcome (Judgment) | Selective reporting (Judgment) |
|---|---|---|---|---|---|---|
|
Turner 2012[27] |
Randomised using a random computer sequence generation (√) |
No information (?) |
Attempted blinding as ALL patients received mailed brochures about heart disease. |
Clinical outcomes (changes in 4 year CHD risk, systolic and diastolic blood pressure) assessors were blinded (√) |
85% completed blood pressure assessment and 69% completed CHD risk assessment. No difference between groups. More withdrawals in intervention group (20/136 v 13/144). Multiple imputation for all missing values (√) |
No protocol, description of clinical assessments correspond to outcomes (X) |
| No self report outcomes (√) | ||||||
|
Walker 2011[28] |
Randomised using a random computer sequence generation (√) |
No information (?) |
Attempted blinding as ALL patients received mailed brochures about heart disease. Self report outcomes used (X) |
No blinding (X) |
87% completed outcomes assessments at 12 months. No difference between groups. More withdrawals in control group (3/264 v 2/262). Multiple imputation for all missing values (√) |
No protocol, description of clinical assessments correspond to outcomes (X) |
| Outcomes self-report by telephone. Physiological measures completed using the ‘dry-dot methodology’ involving patient mailing sample to the lab (?) | ||||||
|
Heisler 2010[33] |
Randomised using a random sequence generation (√) |
Centrally (√) |
Blinded patients, research staff and care managers at baseline. Intervention was described as a comparison of 2 diabetes self-management support models to participants. Not clear after baseline (X) |
Only data assessors were blinded (X) |
89% completed HbA1c assessments and 95% completed survey assessments, no differences between groups, justification is provided (√) |
No protocol, description of measures orresponds to outcomes (X) |
|
Dale 2009[9] |
No details about sequence generation – states randomised only (?) |
Opaque sealed envelopes (X) |
Attempted blinding as ALL patients received one telephone call (X) |
Outcomes self report by post (?) |
91% follow up at 6 months (93.3%, 86.4% and 91.8% overall) no reasons given (?) |
Protocol reported diabetes self care activities measure which was not reported in the main trial (X) |
| Physiological measures assessed blinded to group (√) | ||||||
|
Samuel-Hodge 2009[32] |
Cluster randomised. Computer generated random number (√) |
Sequentially numbered sealed envelopes (X) |
No blinding, self report outcome (X) |
HbA1c measures masked to study group (√) |
87% follow up at 8 months, 85% at 12 months, no difference between groups, more withdrawals in intervention group (6/102 v 1/72) (√) |
No protocol, insufficient information (?) |
| Physical activity not clear (?) | ||||||
| FFQ and other psychosocial outcomes by telephone, masked to study group but not clear if it could have been broken (?) | ||||||
|
Parry 2009[34] |
Internet based randomisation service (√) |
Central (√) |
No blinding, self report outcome (X) |
Researchers blinded to group allocation, self reported outcomes, but not clear if could have been broken (?) |
Follow up 94% at 8 weeks, no difference between groups, reasons given (√) |
No protocol, insufficient information (?) |
|
Batik 2008[30] |
Non random assignment of late new participants to control group (X) |
No information (?) |
No blinding, self report outcome (X) |
Outcomes self-report (?) |
No data reported on follow up (?) |
No protocol, insufficient information (?) |
| Physiological measures (?) | ||||||
|
Carroll 2007[31] |
No details about sequence generation, states randomised only (?) |
No information (?) |
No blinding, self report outcome (X) |
Outcomes self-report via telephone (?) |
18.6% attrition, no reasons given (?) |
No protocol, insufficient information (?) |
|
Young 2005[35] |
Post-recruitment block randomisation, stratified by baseline HbA1c using SAS software (√) |
Randomise intervention to control in a ratio of 2:1 (√) |
No information (?) |
No information (?) |
8.2% lost at follow-up, justification is provided, intention to treat analyses (√) |
No information (?) |
| Keyserling 2002[29] | Randomised using random numbers generated using a personal computer (√) | Consequently numbered sealed envelopes containing study group assignments (X) | No blinding, self report outcome (X) | Clinicians were informed of participants group assignment, no more information is provided (X) | 88% and 84% of participants completed the 6th and 12th month follow-up, no differences between groups, justification is provided (√) | Protocol includes self-care, but no outcomes are reported (X) |
Note: Judgment ratings: √ = Low risk of bias; X = High risk of bias; ? = Unclear risk of bias [22].