Allergens

The performance of methods has been evaluated on a blind or independent dataset obtained from Li et al. (15 (link)). Initially this dataset have 664 allergens where allergens obtained from various sources that includes 238 allergens from International union of immunological societies (IUIS) (), 270 from Swiss-Prot's Allergen Index (), 1171 from the biotechnology information for food safety database (BIFS) () and 752 from food allergy research and resource program (FARRP) (). In this dataset no two sequence have identity >95%. We generate an independent dataset of 323 proteins from the dataset of 664 proteins by removing the common allergens present in the dataset of Bjorklund et al. (12 (link)). These 323 proteins were used as independent dataset for evaluating different methods. We also checked the similarity of 323 allergen sequences to 578 sequences used in developing the algorithm using PROSET software (27 ). It was observed that 48 sequences in independent dataset have 95% or more similarity with protein sequences of 578 used for developing methods.

The primary end-points of this study included the allergen-induced LAR on day 84 assessed by LAR AUC_3–7h and LAR%, and the safety and tolerability of daily ecleralimab administration assessed by AEs and SAEs.
Secondary end-points included day 42 and day 84 EAR AUC_0–2h, EAR%, EAR_min and LAR_min, and day 42 LAR AUC_3–7h and LAR%.
Exploratory end-points included sputum eosinophils, F_ENO, methacholine PC₂₀ and blood eosinophil levels.

The sample size was calculated considering 80% power to detect an absolute difference either of 5% between ecleralimab and placebo in LAR AUC_3–7h at day 84 (assuming that the standard deviation was 7.7) or to detect an absolute difference of 8% between ecleralimab and placebo in LAR% (assuming that the standard deviation was 12.7) using a Hochberg procedure with an overall family-wise controlled type 1 error rate of 10% and assuming a correlation coefficient of 0.8 between the two end-points.
Two unblinded interim analyses (IA) were conducted. IA1 included 20 subjects from Cohort 1 (n=10 in ecleralimab and n=10 in placebo) to confirm the sample size assumptions while assessing the variability in time-adjusted area of percentage decrease in FEV₁ AUC_3–7h and maximum percentage decrease in FEV₁ following the AIC in this population. These data were not intended to change the conduct of this study and were not used to stop the study at this time, but could have informed a change in sample size. No pause in enrolment, pending the data from IA1, was required. IA2 included 28 subjects from Cohorts 1 and 2 (n=15 in ecleralimab and n=13 in placebo), and was conducted when 26 subjects had completed day 84. This analysis showed that there were no safety issues, and that the variability for AUC and maximum percentage decrease was slightly less than assumed (the correlation between these was 0.95). The study was terminated when 27 subjects completed the study. As the study met the primary end-point with the 4-mg dose, a decision was made to cancel the planned dose escalation phase.
The efficacy variables were analysed using the efficacy set, which included all subjects who received any study drug and experienced no protocol deviations with relevant impact on efficacy end-points. The two primary end-points, LAR AUC_3–7h and LAR%, compared with placebo were analysed separately using an ANCOVA model for repeated measures. The model included treatment and visit as independent variables, treatment by visit interaction term, and baseline by visit interaction term where baseline values for time-adjusted area of percentage decrease in FEV₁ AUC_3–7h and maximum percentage decrease in in FEV₁ are measured from the AIC at day −14 during the screening period. Model-based estimates of mean (least square estimates) were reported for ecleralimab and placebo groups on day 42 and day 84 along with the treatment difference, two-sided 90% confidence interval for treatment difference and one-sided p-value.
The secondary efficacy end-points were listed by treatment, subjects and visit/time.
EAR AUC_0–2h, EAR%, EAR_min and LAR_min were also analysed using an ANCOVA model for repeated measures. Baseline values for all secondary efficacy end-points were measured from the screening AIC at day −14.
The exploratory efficacy end-points were listed by treatment, subjects and visit/time, and descriptive statistics were provided by treatment and visit/time, as needed. Methacholine PC₂₀ and F_ENO data were log transformed prior to analysis. Comparisons of ecleralimab with placebo on change from baseline at day 41 and day 83 and allergen-induced shift (change from pre-challenge (day 42/43 to day 41 and day 84/85 to day 83) at 7 and 24 h post-challenge) were carried out using an ANCOVA model for repeated measures for sputum eosinophils, F_ENO and methacholine PC₂₀ (24 h post-challenge). Blood eosinophil levels were compared on change from baseline at day 43 and day 71.
The safety analysis set included all subjects who received any study drug. The primary variable for the primary safety end-point was the occurrence of AEs/SAEs. The numbers and percentages of subjects with AEs by maximum severity of AEs were tabulated by body system and MedDRA preferred term with a breakdown by treatment. All the analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).