Number (%) of trials with multiplicity, number of possible meta-analytic standardised mean differences (SMDs), and differences between meta-analytic SMDs for each index meta-analysis (n=31)
| SR ID | Number of trials | Number (%) of trials with multiplicity | Number of meta-analytic SMDs | Largest minus smallest meta-analytic SMD | Index minus median meta-analytic SMD* | Percentile rank of index meta-analytic SMD† |
|---|---|---|---|---|---|---|
| 1 | 2 | 2 (100) | 6 | 0.34 | 0.16 | 0 |
| 2 | 6 | 3 (50) | 24 | 0.09 | −0.03 | 1 |
| 3 | 7 | 3 (43) | 60 | 0.14 | 0.09 | 0.04 |
| 4 | 2 | 2 (100) | 8 | 0.19 | 0.05 | 0.43 |
| 5 | 8 | 3 (38) | 16 | 0.10 | −0.01 | 0.53 |
| 6 | 3 | 1 (33) | 2 | 0 | 0 | 0 |
| 7 | 5 | 1 (20) | 2 | 0.03 | 0.01 | 0 |
| 8 | 20 | 10 (50) | 3072 | 0.15 | −0.05 | 0.95 |
| 9 | 7 | 1 (14) | 3 | 0.04 | −0.002 | 1 |
| 10 | 6 | 1 (17) | 2 | 0.03 | 0.01 | 0 |
| 11 | 5 | 1 (20) | 3 | 0.05 | 0 | 0.5 |
| 12 | 2 | 1 (50) | 4 | 0.03 | −0.02 | 1 |
| 13 | 2 | 1 (50) | 2 | 0.04 | −0.02 | 1 |
| 14 | 5 | 4 (80) | 864 | 0.27 | −0.09 | 0.93 |
| 15 | 13 | 11 (85) | 2 239 488‡ | 0.43 | −0.06 | 0.76 |
| 16 | 3 | 1 (33) | 2 | 0.01 | −0.01 | 1 |
| 17 | 10 | 4 (40) | 36 | 0.03 | 0.001 | 0.37 |
| 18 | 16 | 9 (56) | 34 992‡ | 0.11 | 0.03 | 0.05 |
| 19 | 3 | 1 (33) | 3 | 0.18 | −0.06 | 1 |
| 20 | 9 | 6 (67) | 384 | 0.15 | −0.06 | 0.98 |
| 21 | 5 | 1 (20) | 3 | 0.01 | −0.001 | 1 |
| 22 | 28 | 6 (21) | 2304 | 0.10 | −0.01 | 0.74 |
| 23 | 2 | 2 (100) | 4 | 0.26 | −0.10 | 0.67 |
| 24 | 5 | 3 (60) | 24 | 0.19 | −0.07 | 0.87 |
| 25 | 13 | 10 (77) | 1 843 200‡ | 0.39 | −0.04 | 0.73 |
| 26 | 6 | 2 (33) | 6 | 0.11 | 0.03 | 0.20 |
| 27 | 9 | 2 (22) | 6 | 0.03 | 0 | 0.40 |
| 28 | 3 | 2 (67) | 4 | 0.42 | −0.20 | 1 |
| 29 | 20 | 18 (90) | 1.8 trillion‡ | 0.16 | −0.03 | 0.85 |
| 30 | 21 | 4 (19) | 576 | 0.03 | 0.01 | 0.01 |
| 31 | 4 | 2 (50) | 576 | 0.28 | −0.12 | 1 |
| Median (IQR) | 6 (3, 10) | 50% (22%, 67%) | 8 (3, 576) | 0.11 (0.03, 0.19) | −0.01 (−0.05, 0.01) | 0.74 (0.20, 1) |
*Differences less than zero indicate that the index meta-analysis SMD is more favourable to the intervention compared with the median meta-analytic SMD.
†When the percentile rank is 1, the index meta-analytic SMD was the most favourable of all possible meta-analytic SMDs.
‡For these meta-analyses, we generated a sampling distribution of 5000 meta-analytic SMDs. Each meta-analytic SMD was created by randomly selecting (with equal probability) an effect estimate for inclusion from each trial, and meta-analysing the chosen effects.