Table 2.
Proportion of participants who achieved treatment success (three or fewer blisters) at 6 weeks by modified intention-to-treat and per-protocol analyses, and subgroup analysis of baseline disease severity and treatment effect by modified intention-to-treat analysis
| mITT population*(all severities†) | Per-protocol population‡ |
mITT population* |
|||
|---|---|---|---|---|---|
| Mild disease† | Moderate disease† | Severe disease† | |||
| Prednisolone (n/N, %) | 92/101 (91%) | 84/91 (92%) | 30/31 (97%) | 41/42 (98%) | 21/28 (75%) |
| Doxycycline (n/N, %) | 83/112 (74%) | 58/78 (74%) | 28/37 (76%) | 36/46 (78%) | 19/29 (66%) |
| Adjusted difference in proportions (%, 90% CI) | 18·6% (11·1 to 26·1)§ | 18·7% (9·8 to 27·6)§ | 21·1% (7·7 to 34·5)¶ | 19·3% (7·7 to 31·0)¶ | 9·5% (−10·4 to 29·4)¶ |
| Unadjusted difference in proportions (%, 90% CI) | 17·0% (8·7 to 25·2) | 17·9% (8·6 to 27·3) | 21·1% (8·4 to 33·8) | 19·4% (8·6 to 30·1) | 9·5% (−10·3 to 29·3) |
| pinteraction‖ | .. | .. | .. | 0·874 | 0·431 |
For doxycycline to be considered non-inferior to the control treatment, the upper bound of the 90% CI should fall below 37%. mITT=modified intention to treat.
All randomised patients who had a blister count at 6 weeks.
Baseline severity; mild=<10 blisters at baseline; moderate=10–30 blisters at baseline; severe=>30 blisters at baseline.
Those participants who before their 6 week visit had not increased the dose of their allocated treatment, changed treatment, or added a new treatment to their allocated treatment (for a reason other than for treatment failure or success), used topical steroids between visit weeks 3 and 6, or missed more than 3 consecutive days of treatment.
Estimates are from a regression model adjusted for baseline severity of bullous pemphigoid and Karnofsky score; age was omitted from the model due to model non-convergence when present.
Estimates are from a regression model containing an interaction between treatment group and baseline severity of disease, also adjusted for Karnofsky score; age was omitted from the model due to model non-convergence when present.
A post-hoc test for interaction was done comparing the mild disease group with both the moderate and the severe disease groups for the adjusted model; a global interaction test gave a pinteraction of 0·709.