Table 2.
Random sequence generation (selection bias) | Allocation concealment (selection bias) | Blinding of outcome assessment (detection bias) (patient-reported outcomes) | Incomplete outcome data addressed (attrition bias) | Lack of selective reporting (reporting bias) | Lack of other sources of bias | Our evaluation | |
---|---|---|---|---|---|---|---|
Bruun-Olsen et al. 2013 [29] | Yes | Yes | Yes | Yes. 6 (2:4) not followed up | Yes | Yes | Good |
Evgeniadis et al. 2008 [19] | Yes | Yes | Yes | Uneven ITT loss to follow up (37.5% intervention and 20% control) | Yes | Yes | Possible bias due to large and uneven losses to follow up |
Frost et al. 2002 [17] | Yes | Not clear | Yes | Uneven loss to follow up (intervention 30%, control 54%) | Yes | Yes | Possible bias due to large and uneven losses to follow up |
Fung et al. 2012 [27] | Yes | Yes | Yes | Yes | Yes | Yes | Good |
Harmer et al. 2009 [30] | Yes | Yes | Yes (mainly) | Yes. ITT, small losses to follow up | Yes | Yes | Good |
Kauppila et al. 2010 [13] | Yes | Probably adequate | No | Yes. Losses to follow up: intervention 18%; control 7%. However patients with incomplete data included in authors’ analyses | Yes | Baseline differences in prevalence of comorbidities and WOMAC score. | Possible risk of bias due to uneven losses to follow up |
Kramer et al. 2003 [25]. Also data from Minns Lowe 2007 [8] | Not described | Not described | Yes | “Medical issue” losses to follow up differed between groups (7.5% in clinic and 15% in home-based groups) | Yes | Yes. ITT analysis reported as well as per-protocol | Possible risk of bias due to uneven losses to follow up between groups |
Liebs et al. 2010 [28] | Yes | Yes | Yes | 11.8% intervention and 18.9% control patients lost to 3 month follow up | Yes | Yes | Possible risk of bias due to uneven losses to follow up |
Madsen et al. 2013 [24] | Yes | Yes | Yes | 10% intervention and 20% control group lost to follow up. Analysis of change scores | Yes | Yes | Possible risk of bias due to uneven losses to follow up |
Minns Lowe et al. 2012 [20] | Yes | Yes | Yes | Yes, low losses to follow up at 12 months | Yes | Yes | Good |
Mitchell et al. 2005 [21] | Yes | Yes | Self-completed questionnaires | Yes | Yes | Randomisation before surgery with pre-surgical intervention component. Surgery cancelled for 24 patients | Good |
Mockford et al. 2008 [14] | Yes | Yes | Yes | 4.7% patients excluded from analysis as lost to follow up | Yes | Yes | Good |
Moffet et al. 2004 [18] | Yes | Yes | Yes | Yes. Uneven loss to follow up at 12 months (intervention 0%, control 20.5%) | Yes | Yes | Good Possible risk of bias for 12 month outcomes |
Monticone et al. 2013 [16] | Yes | Yes | Yes | Yes | Yes | Yes | Good |
Piqueras et al. 2013 [22] | Yes | Yes | Yes | Yes | Yes | Yes | Good |
Piva et al. 2010 [26] | Yes | Yes | Yes | 22.7% control and 14.3% intervention patients lost to follow up | Yes | Yes | Reasonable |
Rajan et al. 2004 [15] | Yes | Not described | Yes | 5.1% intervention and 1.6% control patients lost to follow up | Yes | Yes | Good |
Tousignant et al. 2011 [23] | Yes | Yes | Yes | Similar losses to follow up between groups (intervention 12.5%, control 16.7%) | Yes | 3/24 randomised to control withdrew due to knowledge of group allocation | Reasonable |