Table 2.
Ratio of relative effect measures from observational studies and relative effect measures from RCTs (e.g., RRobs/RRrct): (a) among 74 pairs of pooled estimates, (b) with only one endpoint per review included, and (c) with studies reported in multiple reviews excluded
| Full sample | One endpoint per review | Studies reported in multiple reviews excluded | ||||
|---|---|---|---|---|---|---|
| Proportion | % | Proportion | % | Proportion | % | |
| Ratio > 1*a | 31/74 | 41.9 | 12/29 | 41.4 | 24/65 | 36.9 |
| Ratio < 1*b | 42/74 | 56.8 | 17/29 | 58.6 | 40/65 | 61.5 |
| Ratio = 1* | 1/74 | 1.4 | 0/29 | 0.0 | 1/65 | 1.5 |
| Extreme difference (ratio > 1.43) | 12/74 | 16.2 | 5/29 | 17.2 | 8/65 | 12.3 |
| Extreme difference (ratio < 0.7) | 20/74 | 27.0 | 8/29 | 27.6 | 19/65 | 29.2 |
| Absence of an extreme difference (0.7 ≤ ratio ≤ 1.43) | 42/74 | 56.8 | 16/29 | 55.2 | 38/65 | 58.5 |
*Does not account for direction of effect
aRelative effect larger in observational studies
bRelative effect larger in RCTs