INEXAS: A Phase 2 Randomized Trial of On‐demand Inhaled Interferon Beta‐1a in Severe Asthmatics

Abstract Background Upper respiratory tract infections (URTIs) are important triggers for asthma exacerbations. We hypothesized that inhalation of the anti‐viral cytokine, interferon (IFN)‐β, during URTI, could prevent these exacerbations. Objective To evaluate the efficacy of on‐demand inhaled IFN‐β1a (AZD9412) to prevent severe asthma exacerbations following symptomatic URTI. Methods This was a randomized, double‐blind, placebo‐controlled trial in which patients with severe asthma (GINA 4‐5; n = 121) reporting URTI symptoms were randomized to 14 days of once‐daily nebulized AZD9412 or placebo. The primary endpoint was severe exacerbations during treatment. Secondary endpoints included 6‐item asthma control questionnaire (ACQ‐6) and lung function. Exploratory biomarkers included IFN‐response markers in serum and sputum, blood leucocyte counts and serum inflammatory cytokines. Results Following a pre‐planned interim analysis, the trial was terminated early due to an unexpectedly low exacerbation rate. Asthma worsenings were generally mild and tended to peak at randomization, possibly contributing to the lack of benefit of AZD9412 on other asthma endpoints. Numerically, AZD9412 did not reduce severe exacerbation rate, ACQ‐6, asthma symptom scores or reliever medication use. AZD9412 improved lung function (morning peak expiratory flow; mPEF) by 19.7 L/min. Exploratory post hoc analyses indicated a greater mPEF improvement by AZD9412 in patients with high blood eosinophils (>0.3 × 109/L) at screening and low serum interleukin‐18 relative change at pre‐treatment baseline. Pharmacodynamic effect of AZD9412 was confirmed using IFN‐response markers. Conclusions & Clinical Relevance Colds did not have the impact on asthma patients that was expected and, due to the low exacerbation rate, the trial was stopped early. On‐demand AZD9412 treatment did not numerically reduce the number of exacerbations, but did attenuate URTI‐induced worsening of mPEF. Severe asthma patients with high blood eosinophils or low serum interleukin‐18 response are potential subgroups for further investigation of inhaled IFN‐β1a.


| INTRODUC TI ON
Upper respiratory tract infections (URTIs) are known to be a major risk factor for asthma exacerbations. Up to 95% of asthma exacerbations are associated with the detection of viruses in respiratory secretions, with human rhinoviruses being the most common. 1,2 Thus, URTIs are a significant cause of morbidity and healthcare burden within the asthma population and there is an unmet need for therapeutics which prevent such infections from triggering exacerbations.
The mechanisms by which URTIs trigger exacerbations are poorly understood. One hypothesis is that asthmatic patients have impaired innate anti-viral immunity. Several studies have reported evidence of delayed or deficient type I and/or type III interferon (IFN) response to virus infection in cells from asthmatic patients compared to healthy controls. 3,4 Wark et al showed impaired IFN-β responses in rhinovirus-infected asthmatic bronchial epithelial cells were associated with increased rhinovirus (RV) replication, which returned to normal levels after addition of exogenous IFN-β. 5 However, many other reports have failed to demonstrate this IFN deficiency (as reviewed by Edwards et al 3 ). More recently, IFN impairment has been observed in a subgroup of patients with severe, therapy-resistant atopic asthma but not in patients with well-controlled asthma. 6,7 The above findings led to the hypothesis that exogenous IFN-β could be an effective treatment for the prevention of exacerbations triggered by URTI. Recombinant IFN-β1a was evaluated in a previous study as an inhaled, on-demand therapy for the prevention of asthma worsening following cold or flu symptoms, in a randomized, placebo-controlled trial (NCT01126177 8 ;). Although the primary endpoint, change in 6-item asthma control questionnaire (ACQ-6) score, was not met in the whole cohort, a planned subgroup analysis showed significant benefit of IFN-β1a in patients with severe, difficult-to-treat asthma, both on the primary endpoint and lung function, in particular morning peak expiratory flow (mPEF). 8 To confirm the results of the previous positive findings in the severe asthma subgroup, we performed a randomized, placebo-controlled trial of on-demand inhaled IFN-β1a (hereafter AZD9412) in severe asthmatics. We hypothesized that inhaled AZD9412 would reduce the rate of virally triggered severe exacerbations and hence selected this to be the primary endpoint.
Some of the results of this trial have been previously reported in the form of an abstract. 9

| ME THODS
Further detail on methods can be found in Supporting Information.

| Study design
This was a randomized, double-blind, placebo-controlled trial evaluating the effect of inhaled AZD9412 on severe exacerbations upon URTI ( Figure 1). Asthma patients (GINA steps 4-5 10 ), with a 24-month history of ≥2 severe exacerbations related to URTI (1 within the last 12 months), were screened and recruited into a pre-treatment waiting phase, during which they continued their previous treatment regimen (maintenance treatment with medium-to-high dose ICS [>250 μg fluticasone total daily dose] and a second controller medication). Patients were equipped with a home spirometer and a smartphone for questionnaires. During the pre-treatment phase, while waiting for URTI symptoms to occur, a range of baseline assessments was completed, including lung function, blood, sputum, nasal lavage and urine samples for biomarkers, ECG recordings, ACQ-6 and Asthma Quality of Life Questionnaire (AQLQ). On 4 consecutive days every month, patients were asked to complete a questionnaire with 10 questions on symptoms of colds and flu, in order to determine baseline. Patients were asked daily via an eDiary On-demand AZD9412 treatment did not numerically reduce the number of exacerbations, but did attenuate URTI-induced worsening of mPEF. Severe asthma patients with high blood eosinophils or low serum interleukin-18 response are potential subgroups for further investigation of inhaled IFN-β1a. device if they thought they were developing a URTI (common cold or influenza). When a patient first reported onset of relevant symptoms (≥2 of sore throat, nasal symptoms [runny and/or blocked nose] different than normal, feeling feverish), they were randomized, as soon as possible but no later than 48 hours after the onset of symptoms, to a 14-day course of 6 million units of once-daily nebulized (iNeb, Philips) AZD9412 or placebo (both Rentschler Biopharma). From the onset of symptoms, patients rated their severity according to the modified Jackson cold score questionnaire on the eDiary each morning during the treatment and follow-up periods, to clinically verify colds. 11,12 The following common cold/flu symptoms were included: sore throat, runny nose, sneeze, nasal congestion (blocked or stuffy nose), malaise (tiredness), fever (feverish/chills), headache, hoarseness, earaches and cough. Each symptom was rated on a scale from 0 = no symptoms, 1 = mild, 2 = moderate, 3 = severe.
Primary endpoint was the occurrence of severe exacerbations following onset of URTI for 14 days from start of treatment, compared with placebo. Secondary endpoints included occurrence of severe or moderate exacerbations (defined in Supporting Information) for up to 30 days from start of treatment, and changes from baseline in the 6-item asthma control questionnaire (ACQ-6), 13 forced expiratory volume in 1 second (FEV 1 ), PEF, reliever medication use, Asthma Symptom Diary scores, safety and tolerability. CompEx (Composite endpoint for Exacerbations) 14 was also evaluated. CompEx is a novel composite endpoint that captures clinically relevant asthma worsening episodes, based on a combination of asthma worsening diary events (morning and evening peak expiratory flow, symptoms and rescue medication use) plus severe exacerbation events. Patients attended the clinic every 3-4 days during treatment and at days 17 and 30, for clinical examination and collection of blood and expectorated sputum.
A pre-planned, un-blinded, administrative interim analysis of the primary efficacy outcome was conducted when 50% of patients had completed the treatment phase, by sponsor personnel who were not involved in the conduct of the study. This was intended to facilitate investment decisions by the study sponsor. Blinded interim monitoring of the severe exacerbation event rate was also performed during the study to calculate, based on the number of events occurring in the two treatment arms combined, whether or not the sample size should be re-estimated.

| Exploratory biomarkers
The presence of 21 pathogens in nasal swab and sputum from the first 7 days of treatment was determined using the Respiratory Sputum mRNA expression of CXCL10, GBP1, Mx1, OAS1 and IFIT2 was quantified by qRT-PCR. Expression was normalized to the geometric mean of 3 housekeeping genes (PUM1, ACTB and HPRT1).

| Statistical methods
The primary endpoint was analysed using a log-binomial regression Thirty per cent of the patients in the placebo arm were expected to experience a severe exacerbation during treatment, based on previous studies in a similar patient population. N = 97 evaluable patients in each arm were required to provide 80% power to discover a relative risk reduction of 55% between the AZD9412 and placebo treatment arms at a significance level of 5%. Based on these calculations, 220 patients were scheduled to be randomized.

| Patients
Three-hundred and forty-nine patients were enrolled in the study, of whom 121 were randomized (ITT population). Patients were recruited in the following countries: Argentina (n = 48), United Kingdom (n = 31), South Korea (n = 21), Spain (n = 8), Australia (n = 6), France (n = 4) and Colombia (n = 3). Sixty-one and 60 patients received AZD9412 and placebo, respectively ( Figure 2). However, the trial was stopped following the pre-planned interim analysis due to an unexpectedly low exacerbation rate and corresponding lack of evidence for differential response on the primary endpoint. The resulting reduction in patient numbers substantially reduced the statistical power of the study analyses. Demographics and patient characteristics at baseline are described in Table 1. There were no major differences between the active and placebo arms. Of the 228 patients not randomized, 214 did not meet the main randomization criteria (development of URTI symptoms) before study termination and 14 withdrew for other reasons ( Figure 2). For those 121 patients who were randomized, the median waiting time from screening to randomization was 42 days (range 8-224 days). Of the 121 randomized patients, 117 completed the study. Greater than 80% adherence to the study medication (ie 12 or more of the 14 once-daily doses were administered) was achieved in 93.4% and 96.7% of the active and placebo arms, respectively.

| Exacerbations and CompEx in AZD9412 vs placebo
The number of patients with a severe exacerbation between days 1 and 14 was 7 (11.5%) and 5 (8.3%) in the active and placebo arm, respectively, with a rate ratio of 1.29 (95% CI 0.43 to 3.85, P = .645; Table 2). The proportion of patients with a severe exacerbation between days 1 and 30 was also similar between the two arms, as was time to severe exacerbation (data not shown).
Similar to the findings on exacerbations, no statistically significant difference was found in the proportion of patients with a CompEx event between the active and placebo arms, although the reduced number of patients limited statistical power (Table 3).

| Lung function in AZD9412 versus placebo
There was a tendency towards an increase in mean mPEF in both arms over the treatment period ( Figure 3). The average AUC mPEF was statistically significantly greater in the active arm over days 1 to 7 compared with that seen in the placebo arm, LS mean of 21.9 L/min vs 2.1 L/min, with a LS mean difference of 19.7 L/min (95% CI 4.8 to 34.6, P = .01; Table 4). Likewise, there was an increase in average AUC mPEF in the active arm over days 1 to 14 compared with a decrease in the placebo arm [9.6 L/ min vs -7.6 L/min, LS mean difference of 17.2 L/min (95% CI -0.5 to 34.9, P = .06; Table 4)]. There was no significant difference in mean percentage change from baseline FEV 1 at any time point (Table S1).

| ACQ-6, asthma symptom scores and reliever medication use
There were no significant differences in ACQ-6, asthma symptom scores or reliever medication use between AZD9412 and placebo ( Figure S1, Tables S2 -S4). The highest levels for asthma symptom scores and reliever medication use, from randomization to follow-up, were observed at randomization, with a steady decline throughout the treatment period ( Figure 4).

| Pharmacodynamic effect of AZD9412
Compared to screening, serum concentrations of the IFN-response biomarker, CXCL10, increased in both placebo and AZD9412 arms at day 1, when patients had reported URTI symptoms, immediately before the first dose ( Figure 5A). During the treatment period, patients on AZD9412 maintained significantly higher CXCL10 concentrations compared to placebo (P = .016, Figure 5A). By 3 days after end of treatment (visit 7), CXCL10 in the AZD9412 arm had returned to screening levels.
In sputum, the time courses of mRNA expression for 5 interfer- were similar to serum CXCL10 protein, with a modest increase around day 1 in both arms compared to screening ( Figure 5B).
Throughout the treatment period, levels of all 5 ISG mRNA were higher in the AZD9412 arm compared to placebo, and this was statistically significant for all but GBP1 (P < .001). Unlike serum CXCL10 protein, in the AZD9412 arm all 5 ISG mRNA increased further from day 1 to day 2 and peaked later (around day 3 for GBP1 and day 7-10 for the others).

| Pre-defined subgroup analyses
Sub-division of the cohort based on northern or southern hemisphere or clinically verified colds (92% of the ITT population) showed similar results to the main ITT population (Tables S5 and S6). Patients were divided into serum CXCL10-low, medium and high subgroups based on pre-treatment baseline levels, and the results were similar between all 3 subgroups (Table S7A-C).
Overall, 52% of evaluable patients were virus-positive, of whom 31% were rhinovirus-positive (Table S8) (Table S9A and   Table 4). However, as in the ITT population, there was no effect of AZD9412 on any other secondary endpoint (   (Table 5 and Figure 6A-B).
We also aimed to determine whether a soluble biomarker could be utilized to identify patients with high blood eosinophils in this cohort. Eosinophil-derived neurotoxin (EDN) is a granule protein released from eosinophils upon activation and as such, has been described in the literature as a robust biomarker of activated eosinophils. 15 We therefore measured EDN concentrations in serum from 51 patients at screening. Serum EDN concentrations correlated strongly with baseline blood eosinophil counts (r 2 = .61; Figure 6C).

| Safety and Tolerability
The safety profile of AZD9412 was consistent with previous clinical experience, 9 and no new safety concerns were identified. The proportion of patients reporting adverse events (AEs) was 47.5% in the AZD9412 arm compared with 33.3% in the placebo arm (  TA B L E 5 Morning peak expiratory flow (mPEF), area under the curve (AUC) change from baseline in subgroups defined by four different biomarkers: serum IL-18 fold change, serum TRAIL fold change, eosinophils (10 9 /L) at baseline and neutrophils (10 9 /L) at baseline F I G U R E 6 A, Mean mPEF AUC change (day 1-7) in patients sub-divided on blood eosinophil counts at screening. Patient numbers in low, mid and high subgroups were 9, 32 and 12 for AZD9412 and 19, 27 and 12 for placebo. B, Mean mPEF AUC change (day 1-7) in patients subdivided on mean IL-18 relative change from screening to treatment baseline. Patient numbers in low, mid and high subgroups were n = 14, 29 and 13 for AZD9412, and n = 14, 30 and 13 for placebo. A-B, All = main ITT population. Error bars are 95% confidence intervals. C, Scatterplot of blood eosinophil counts versus EDN (log 10 -scale) based on all patients with both EDN and eosinophil count data at screening (n = 47, excluding one patient whose blood eosinophil count = 0 and 3 patients with missing eosinophil count data). Correlation between EDN and blood eosinophil counts (log 10 -scale) is 0.78 (Pearson's correlation coefficient) mPEF during the first 7 days of treatment that was similar to the previous trial of inhaled IFN-β1a in patients with asthma. 8 The difference of 19.7 L/min, however, is of limited clinical significance. Our post hoc exploratory analyses indicated a greater mPEF improvement in patients with either high blood eosinophils at screening or low serum IL-18 response at the time of URTI symptoms. Finally, asthma worsenings were generally mild and tended to peak at the start of treatment.

| D ISCUSS I ON
We originally aimed to randomize 220 patients based on an expected event rate of 30% in the placebo arm. However, following a pre-planned interim analysis after 50% of patients had been randomized, a low event rate was observed (8% in placebo; see Table 2), with no difference between AZD9412 and placebo.
These findings led to the decision to terminate the study following the randomization of 121 patients. However, it was assessed before termination of the trial that the sample size was large enough to have sufficient power to evaluate key secondary endpoints (ACQ-6 and mPEF). We acknowledge that there were limitations which may explain our over-estimation of the anticipated event rates. Although we used previous clinical experience and epidemiological data to estimate expected event rates, 16 Although our trial replicates these mPEF findings, we saw no effect of AZD9412 on ACQ-6. The previous trial did not report on the rate of exacerbations. Despite the difference in the ACQ-6 outcome, our study population was comparable with the difficult-to-treat subgroup in Djukanovic et al. 8 We used the GINA steps 4-5 classification rather than those of the British Thoracic Society. Although the criteria are similar, there may be subtle differences between them, 10,20 and a documented history of severe exacerbations related to URTI was required for enrolment. It is notable that in our trial, ACQ-6 values were higher at screening and, unlike in the previous trial, did not increase in the placebo arm following URTI symptoms (see Figure S1). it is tempting to speculate that asthmatics with high blood eosinophils or low IL-18 may have impaired innate anti-viral immunity and would therefore be more likely to benefit from IFN-β1a therapy.
However, this hypothesis would require confirmation. In summary, respiratory viral infections did not have the expected impact and, due to the low exacerbation rate, our evaluation of on-demand inhaled AZD9412 versus placebo for the prevention of severe asthma exacerbations following URTI symptoms was stopped early. AZD9412 showed no differential effect on severe exacerbations, but did give rise to an improvement in mPEF, a response which in an exploratory analysis tended to be greater in patients with either high blood eosinophils or low serum IL-18 relative change at pre-treatment baseline. The finding that changes in asthma endpoints were minimal, and had already peaked at randomization, suggests that early detection of asthma worsening, and identifying patients more likely to deteriorate due to respiratory viruses, is essential for optimal efficacy of on-demand IFN-β1a therapy. Our findings should be taken into consideration for the future development of inhaled IFN-β1a.