Symptoms and Laboratory Values as Proxies for Endoscopic and Histologic Clinical Endpoints in Ulcerative Colitis: A Mediation Analysis Based on Upadacitinib Phase 3 Induction Trials

Introduction: We evaluate the extent to which commonly available measures in clinical practice (i.e., signs, symptoms, and laboratory values) mediate endoscopy-based clinical measures. Methods: We analyzed, in the intent-to-treat population from upadacitinib’s (UPA) phase 3 induction trials (U-ACHIEVE Induction, U-ACCOMPLISH in moderately to severely active UC, relationships between deep mucosal healing (DMH), histologic endoscopic mucosal improvement (HEMI), and change from baseline in endoscopic score as outcomes, and signs/symptoms and lab values as mediators (fecal calprotectin, high sensitivity C-reactive protein [hs-CRP], abdominal pain [AP], bowel urgency [BU], and Partial Adapted Mayo Score [PA-Mayo]). The mediated proportion of effect at week (wk) 8 was calculated as the difference in mean outcomes due to a one-unit change in mediators while controlling for baseline characteristics.The effects of mediators at wks 2 and 8 were assessed. Analyses assumed mediators were unrelated to each other and used linear models for all variables. Missing values for DMH and HEMI at wk 8 were imputed via non-responder imputation per trial protocol, while other variables were analyzed as-observed. Standard errors were computed via bootstrap. Results: A total of 878 patients were included; mean (standard deviation [SD]) age was 42.9 (14.4) years and disease duration was 8.0 (7.2) years. At wk 8, 9.0% of patients achieved DMH and 25.7% achieved HEMI, and mean (SD) change from baseline of the endoscopic score was -0.70 (0.97). All outcomes were mediated by PA-Mayo and BU (Table). PA-Mayo mediated around a third of the total effect for all outcomes with similar mediation at wk 2 and 8 (all P< 0.05). The proportion mediated by BU was larger at wk 8 and ranged from one to two thirds of the total effect (all P< 0.05 except for wk 2 for DMH). In contrast, hs-CRP at wk 2 only mediated 2-3% across outcomes (all P< 0.05). The predictive power of the models was similar using wk 2 and 8 mediators and explained 15% (DMH), 40% (HEMI), and 50% (endoscopic score) of variation. Conclusion: Our study found that easily accessible measures such as Partial Adapted Mayo Score and bowel urgency can mediate endoscopy-based endpoints for UC as early as wk 2 of induction, with mediation effects persisting for concurrently assessed signs and symptoms. Our findings suggests that symptom-based measures for UC in clinical practice could be useful proxies for endoscopic and histologic outcomes. Table 1. - Proportion mediated as estimated from mediation analyses for week 8 outcomes using week 2 or week 8 mediators Deep mucosal healingc HEMId Change from baseline in endoscopic score Variance explained Week 2 (R2= 0.15b) Week 8 (R2= 0.15 b) Week 2 (R2= 0.39 b) Week 8 (R2= 0.40 b) Week 2 (R2= 0.50 b) Week 8 (R2= 0.52 b) Bowel urgency, proportion mediated, %a 30.2 66.2* 48.12* 49.4* 37.4* 45.1* Partial Adapted Mayo Score, proportion mediated, %a 33.6* 30.1* 34.6* 33.2* 27.6* 29.1* Abdominal pain, proportion mediated, %a -4.3 -17.4 -13.7 0.4 -5.1 -5.0 hs-CRP (mg/L), proportion mediated, %a 1.8* -0.1 2.6* 1.3 2.3* 1.3 Fecal calprotectin (mg/kg), proportion mediated, %a -0.0 -1.7 -0.0 -1.6 -0.2 -1.9 Abbreviations: HEMI: Histologic Endoscopic Mucosal Improvement, hs-CRP: High sensitivity C-reactive protein.* denotes statistical significance (alpha < 0.05).Note: [a] Proportion mediated denotes the proportion of the mediated effect of changing one unit of the mediator relative to the total effect; for fecal calprotectin, a change of 100 mg/kg was considered.[b] The R2 was calculated for the model regressing the outcome on all control variables and all mediators.[c] Deep mucosal healing was defined as endoscopic score of 0 and Geboes score < 2.0.[d] HEMI was defined as endoscopic score of 0 or 1 and Geboes score ≤ 3.1.