Table 3 Concordance between statistically significant (P < .05) originally reported meta-analytical effect size and the re-estimated effect size based upon trial classification
Fixed effects model | Random effects model | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Â | k | Kappa | (95CI) | Â | Rendered non-significant | Rendered significant | k | Kappa | (95CI) | Rendered non-significant | Rendered significant |
Trial classification | |||||||||||
N > 100a | 113 | 0.68 | (0.63, | 0.72) | 15% | 0% | 0.58 | (0.53, | 0.62) | 20% | 2% |
N > 370a | 84 | 0.47 | (0.41, | 0.52) | 20% | 5% | 0.35 | (0.30, | 0.41) | 26% | 7% |
Excluding pilot/feasibility | 83 | 0.91 | (0.89, | 0.94) | 4% | 0% | 0.81 | (0.77, | 0.84) | 8% | 1% |
Excluding N ≤ 100 | 83 | 0.68 | (0.64, | 0.72) | 14% | 0% | 0.65 | (0.60, | 0.69) | 17% | 1% |