Table 4.
Associations between review characteristics and overall ROBINS-I risk of bias judgements reported in the included systematic reviews
| Odds ratio (95% CI) |
Marginal predicted probability, % (95% CI) |
N studies (reviews) | |||||
|---|---|---|---|---|---|---|---|
| Low vs. higher RoB |
Low/moderate vs. higher RoB |
Critical vs. lower RoB |
Low RoB | Low/moderate RoB | Critical RoB | ||
| Review qualitya | |||||||
| High/Moderate (baseline) | 5 (2, 11) | 31 (19, 46) | 30 (17, 45) | 426 (21) | |||
| Low | 1.9 (0.36, 10.0) | 3.3 (0.78, 14) | 0.19 (0.04, 0.82) | 8 (4, 14) | 46 (35, 58) | 13 (7, 22) | 409 (31) |
| Critically low | 4.7 (1.0, 22) | 7.7 (1.9, 31) | 0.24 (0.06, 0.98) | 13 (8, 20) | 57 (47, 67) | 15 (9, 23) | 504 (41) |
| Review included RCTsa | |||||||
| No (baseline) | 12 (7, 18) | 45 (34, 56) | 19 (13, 28) | 701 (40) | |||
| Yes | 0.59 (0.17, 2.0) | 1.4 (0.47, 4.4) | 0.69 (0.21, 2.2) | 9 (5, 14) | 50 (40, 60) | 16 (10, 23) | 638 (53) |
| RoB assessed in duplicatea | |||||||
| Yes (baseline) | 11 (7, 16) | 47 (38, 55) | 22 (15, 29) | 992 (63) | |||
| No | 0.34 (0.08, 1.5) | 1.3 (0.40, 4.4) | 0.20 (0.05, 0.70) | 6 (2, 12) | 51 (38, 63) | 10 (5, 16) | 347 (30) |
| Industry funding or competing interestsb | |||||||
| No (baseline) | 14 (9, 20) | 50 (42, 58) | 17 (12, 23) | 795 (62) | |||
| Yes | <0.01 (<0.01, 0.07) | 0.48 (0.10, 2.3) | 0.56 (0.10, 3.1) | 0 (0, 3) | 41 (23, 60) | 13 (5, 29) | 231 (13) |
CI, confidence interval; RoB, risk of bias.
A separate generalized ordered logit regression model was fitted for each predictor, with a review-level random intercept.
a
N = 1339 studies, 93 reviews.
b
N = 1026 studies, 75 reviews.