Table 3.
Type I error rate and CIs for standard and augmented binary methods
| Statistical property | Difference |
|
|---|---|---|
| Standard binary (%) | Augmented binary (%) | |
| ACR20 | ||
| Type I error rate (%) | 5.1 | 5.1 |
| Average width of CI | 15.7 | 13.2 |
| Relative percentage reduction in width | 15.8 | |
| ACR50 | ||
| Type I error rate | 4.9 | 5.2 |
| Average width of CI | 11.6 | 9.5 |
| Relative percentage reduction in width | 17.9 | |
| ACR70 | ||
| Type I error rate | 4.6 | 5.0 |
| Average width of CI | 7.5 | 4.7 |
| Relative percentage reduction in width | 38.1 | |
| DAS28 <3.2 | ||
| Type I error rate | 5.9 | 5.2 |
| Average width of CI | 12.1 | 8.8 |
| Relative percentage reduction in width | 27.4 | |
| DAS28 <2.6 | ||
| Type I error rate | 5.0 | 5.3 |
| Average width of CI | 9.0 | 5.4 |
| Relative percentage reduction in width | 40.1 | |
For each outcome, 5000 simulation replicates are used (Monte-Carlo error ±0.6%). The average width of the CIs is on the logarithmic scale for the risk ratio and odds ratio outcomes.