Table 4.
Multivariable Logistic Regression Models of Baseline and Week 4 Characteristics Associated with Week 12 CS-free remission for all therapies and additional therapy/colectomy for patients initially treated with IV corticosteroids
| CS-Free Remission, all patients | CS-Free Remission by Initial Treatment | Additional Therapy/Colectomy | |||
|---|---|---|---|---|---|
|
| |||||
| Odds Ratio (95% CI) p-value |
Total (N=409)# | 5-ASA (N=129) | Oral CS (N=139) | IV CS (N=141) | IV CS only (N=141) |
|
| |||||
| Model sample size (% of total N) | n=403 (99%) | n=116 (90%) b | n=139 (100%) b | n=119 (84%) | n=119 (84%) |
|
| |||||
| Number of events (% of model n) | 140 (35%) | 57 (49%) | 47 (34%) | 26 (22%) | 42 (35%) |
|
| |||||
| Baseline predictors: | |||||
|
| |||||
| Lower PUCAI |
PUCAI <35: 2.44 (1.41, 4.22) 0.001 |
- |
PUCAI < 45: 4.38 (1.81, 10.60) 0.0011 |
- | - |
|
| |||||
| Total Mayo score ≥11 | - | - | - | - | 2.59 (0.93, 7.21) 0.068 |
|
| |||||
| Higher albumin per 1g/dL increase (interaction with age)a | For Age < 12: 4.05 (1.90, 8.64) 0.0003 For Age ≥ 12: 1.13 (0.74, 1.71) 0.57 |
- | - | - | - |
|
| |||||
| Hemoglobin ≥12 g/dL | - | 2.19 (0.96, 4.97) 0.062 |
- | - | - |
|
| |||||
| Rectal biopsy eosinophil peak count ≤32/hpf | - | - | - | - | 4.55 (1.62, 12.78) 0.004 |
|
| |||||
| Rectal biopsy surface villiform changes | - | - | - |
No changes: 2.71 (0.97, 7.56) 0.057 |
Changes: 3.05 (1.09, 8.56) 0.034 |
|
| |||||
| Week 4 Remission | 6.26 (3.79, 10.35) < 0.0001 |
3.69 (1.67, 8.15) 0.0013 |
8.02 (3.11, 20.70) <0.0001 |
7.48 (2.67, 20.96) 0.0001 |
No Remission: 30.28 (6.36, 144.20) < 0.0001 |
|
| |||||
| AUC (95% CI) | 0.79 (0.74, 0.84) | 0.70 (0.61, 0.79) | 0.78 (0.71, 0.86) | 0.77 (0.67, 0.88) | 0.89 (0.83, 0.95) |
| Adjusted AUC | 0.78 | 0.70 | 0.78 | 0.77 | 0.88 |
| R2 | 0.31 | 0.18 | 0.32 | 0.26 | 0.54 |
|
| |||||
| Sensitivity (95% CI) | 41% (33%, 50%) | 54% (41%, 68%) | 40% (26%, 56%) | 62% (41%, 80%) | 71% (55%, 84%) |
|
| |||||
| Specificity (95% CI) | 89% (85%, 92%) | 61% (47%, 73%) | 93% (86%, 98%) | 86% (77%, 92%) | 86% (76%, 93%) |
|
| |||||
| PPV (95% CI) | 67% (56%, 76%) | 57% (43%, 71%) | 76% (55%, 91%) | 55% (36%, 74%) | 73% (57%, 86%) |
|
| |||||
| NPV (95% CI) | 74% (69%, 79%) | 58% (45%, 70%) | 75% (66%, 83%) | 89% (81%, 95%) | 85% (75%, 92%) |
N is the number Evaluable at week 12 and with no protocol violations.
Adjusted AUC = AUC with bootstrap adjustment for optimism of model estimation, PPV=positive predictive value, NPV = negative predictive value
For the full cohort, p-value for interaction between age<12 and albumin: p= 0.004.
Further models for the full cohort identified an interaction between mild baseline PUCAI and week 4 remission (p=0.05), indicating that the impact of week 4 remission on week 12 remission is less strong for participants with a mild baseline PUCAI. The interaction was not included in the model because there was no improvement in the model AUC.
Potential covariates were excluded from the model based on internal validation as follows: 5-ASA and Oral CS groups: ANCA positive (due to clinically inconsistent between subgroups and across assessment times); additional therapy/colectomy of IV CS group: OmpC<12 (due to lack of association in the multiple imputation model).