Trends over time in proportions of trials classified high, unclear or low for each risk of bias domain
| Risk of bias domain | Pre-2002 N=37 | 2002–2006 N=46 | 2007–2010 N=59 | p Value* | p Value† |
|---|---|---|---|---|---|
| Was the allocation sequence adequately generated? | 0.41 | 0.43 | |||
| Low | 11 (30%) | 19 (41%) | 25 (42%) | ||
| Unclear | 24 (65%) | 25 (55%) | 33 (56%) | ||
| High | 2 (5%) | 2 (4%) | 1 (2%) | ||
| Was the allocation adequately concealed? | 1.00 | 0.82 | |||
| Low | 15 (40%) | 20 (44%) | 25 (42%) | ||
| Unclear | 21 (57%) | 25 (54%) | 32 (54%) | ||
| High | 1 (3%) | 1 (2%) | 2 (4%) | ||
| Were baseline outcomes similar? | 0.87 | 0.20 | |||
| Low | 31 (84%) | 39 (85%) | 44 (75%) | ||
| Unclear | 2 (5%) | 3 (6%) | 8 (13%) | ||
| High | 4 (11%) | 4 (9%) | 7 (12%) | ||
| Were baseline characteristics similar? | 0.16 | 0.57 | |||
| Low | 30 (81%) | 34 (74%) | 43 (73%) | ||
| Unclear | 3 (8%) | 6 (13%) | 3 (5%) | ||
| High | 4 (11%) | 6 (13%) | 13 (22%) | ||
| Were incomplete outcome data adequately addressed? | 0.17 | 0.70 | |||
| Low | 29 (78%) | 33 (72%) | 38 (64%) | ||
| Unclear | 3 (8%) | 8 (17%) | 7 (12%) | ||
| High | 5 (14%) | 5 (11%) | 14 (24%) | ||
| Was knowledge of the allocated interventions prevented? | 0.44 | 0.61 | |||
| Low | 32 (87%) | 38 (83%) | 46 (78%) | ||
| Unclear | 3 (8%) | 5 (11%) | 7 (12%) | ||
| High | 2 (5%) | 3 (6%) | 6 (10%) | ||
| Was the study protected against contamination? | 0.54 | 0.78 | |||
| Low | 25 (68%) | 23 (50%) | 35 (59%) | ||
| Unclear | 10 (27%) | 21 (46%) | 19 (32%) | ||
| High | 2 (5%) | 2 (4%) | 5 (9%) | ||
| Was the study free from selective outcome reporting? | 0.84 | 1.00 | |||
| Low | 3 (8%) | 4 (9%) | 6 (10%) | ||
| Unclear | 32 (87%) | 37 (80%) | 50 (85%) | ||
| High | 2 (5%) | 5 (11%) | 3 (5%) | ||
| Was the study free from other risks of bias? | 0.58 | 0.58 | |||
| Low | 27 (73%) | 30 (65%) | 39 (66%) | ||
| Unclear | 10 (27%) | 16 (35%) | 20 (34%) | ||
| High | 0 | 0 | 0 |
*Exact Cochran-Armitage test for high versus low or unclear risk of bias in each domain except the last domain which was analysed as low versus high or unclear due to absence of studies with high risk of bias.
†Exact Cochran-Armitage test for reported (high or low risk of bias) or unreported (unclear risk of bias) in each domain.