Table I.
Power to detect improvements in live birth rate of different sizes for randomised controlled trials and meta-analyses included in the review.
| 5-pp LBR improvement | 10-pp LBR improvement | 15-pp LBR improvement | 20-pp LBR improvement | ||
|---|---|---|---|---|---|
| Largest RCT in comparison N = 199 | Median power | 13% | 34% | 63% | 85% |
| IQR | 8–21% | 18–59% | 34–88% | 52–99% | |
| Range | 5–69% | 6–100% | 8–100% | 10–100% | |
| Proportion with 90% (80%) power | 0% (0%) | 6% (10%) | 23% (33%) | 44% (55%) | |
| Meta-analysis for comparison N = 199 | Median power | 16% | 45% | 75% | 93% |
| IQR | 9–33% | 21–83% | 40–99% | 60–100% | |
| Range | 5–100% | 6–100% | 8–100% | 10–100% | |
| Proportion with 90% (80%) power | 1% (2%) | 19% (27%) | 39% (46%) | 53% (60%) |
For example, having a power of 80% to detect an improvement in live birth rate (LBR) of a given size (e.g. 5 pp) means that a study would have an 80% chance of a statistically significant result (P < 0.05) under the condition that the effect of treatment on LBR was of this particular size (5 pp) and a standard statistical analysis was used (chi-squared test). RCT, randomised controlled trial; IQR, interquartile range; pp, percentage points.