Introduction

Chronic cough (CC), estimated to affect around 10% of the global adult population [1], is defined as a cough lasting ≥ 8 weeks [2]. CC with unexplained underlying etiology or CC that is unresponsive to conventional treatment are jointly referred to here as refractory CC (RCC) [2]. RCC can have a detrimental impact on patients’ quality of life (QoL) [3] and mental health [4, 5] and results in significant economic burden, with patients experiencing repeated treatment failures and delayed diagnosis [4]. There are no approved drugs for RCC in countries other than Japan [6] and Switzerland [7], resulting in widespread use of off-label treatment options with limited efficacy and a poor safety profile, and non-pharmacologic interventions [4]. There is therefore a large unmet clinical need for efficacious, well-tolerated therapies.

Neuronal hypersensitivity is implicated in the pathogenesis of RCC [8]. Patients with RCC have increased cough reflex sensitivity, which may result from vagal nerve hypersensitivity or changes in the central nervous system projections and central sensitization as presumed underlying mechanisms [8]. P2X3 receptors are thought to play an important role in sensory neural dysregulation associated with RCC [9, 10]. Preclinical studies have shown that P2X2/3 receptors can regulate afferent sensory adenosine triphosphate-mediated signaling in the vagus nerve [9]. Clinical trials of the P2X3 receptor antagonist gefapixant showed efficacy in objective and subjective measures of cough in patients with RCC [11,12,13]. However, substantial taste-related tolerability issues [11,12,13], attributed to the block of P2X2/3 receptors on nerves innervating taste buds [14], may limit acceptance of gefapixant by patients.

Eliapixant is a potent P2X3 receptor antagonist with a good tolerability profile in healthy subjects, and high selectivity over the P2X2/3 receptor in vitro, potentially resulting in fewer off-target effects [15,16,17]. In a phase 2a study, eliapixant significantly reduced cough frequency and severity in patients with RCC, with a lower rate of taste-related side effects than those observed with therapeutic doses of gefapixant [18]. The aim of the phase 2b PAGANINI study was to identify the optimal dose of eliapixant in patients with RCC, to further assess efficacy, and to characterize the safety and tolerability profile of eliapixant over 12 weeks.

Methods

Study Design

PAGANINI (ClinicalTrials.gov NCT04562155) was a randomized, double-blind, parallel-group, placebo-controlled, dose-finding efficacy and safety study conducted at 99 centers in 19 countries (see Supplementary Methods for more details). The study consisted of a 14-day screening period, 12 weeks of randomized treatment, and a 30-day safety follow-up (Supplementary Fig. S1). The study protocol and statistical analysis plan are available on ClinicalTrials.gov.

Eligible participants were centrally randomized 1:1:1:1 by the sponsor using block randomization to receive one of three oral doses of twice-daily eliapixant (25 mg, 75 mg, or 150 mg; Bayer AG, Berlin, Germany) or placebo using Interactive Response Technology (IRT version 2.1; Suvoda, USA), stratified by region. To maintain blinding, tablets containing eliapixant and placebo were identical in size, color, and shape.

Participants

Adults aged ≥ 18 years with RCC lasting ≥ 12 months, with persistent cough for ≥ 8 weeks before screening, and with cough severity ≥ 40 mm measured on a 100 mm visual analog scale (VAS) at screening, were enrolled by the investigators. Full inclusion and exclusion criteria are in the Supplementary Methods.

Procedures

Using an ambulatory cough recording device (VitaloJAK, Vitalograph, Ireland [19]), 24-h cough count monitoring was performed at every visit to Week 12 (see Supplementary Methods for more details). Participants completed the cough severity VAS [20] daily and the Leicester Cough Questionnaire (LCQ) [21] at all visits (see Supplementary Methods for more details). Adverse events (AEs) and other safety outcomes were evaluated throughout the study and at follow-up.

Outcomes

The primary efficacy endpoint was change from baseline in 24-h cough count after 12 weeks of intervention. Secondary efficacy endpoints included: the percentage of participants with ≥ 30% reduction from baseline 24-h cough count after 12 weeks; change from baseline 24-h cough count after 2, 4, and 8 weeks; change from baseline awake cough count per hour after 2, 4, 8, and 12 weeks; change from baseline cough severity after 12 weeks measured by the cough severity VAS; the percentage of participants with ≥ 30-scale unit reduction from baseline after 12 weeks measured by the cough severity VAS [22]; change from baseline cough-related QoL after 12 weeks measured by the LCQ; and the percentage of participants with ≥ 1.3-point increase from baseline after 12 weeks measured by LCQ total score [23].

Treatment-emergent AEs and serious AEs (SAEs) were recorded according to the Medical Dictionary for Regulatory Activities version 24.0. Additional safety assessments are described in the Supplementary Methods. At the study end, participants who spontaneously reported a taste-related AE completed an assessment on taste disturbances.

Statistical Analysis

A multiple comparison procedure modeling (MCP-Mod) approach [24] was used as the prespecified analysis of the primary efficacy endpoint. As PAGANINI was a phase 2b dose-finding study, the MCP-Mod approach was used because it is a well-accepted method for dose finding that efficiently uses the available data better than traditional pairwise comparisons [25, 26]. The MCP-Mod approach enables the estimation of a dose response and the selection of an optimum dose for further phase 3 trials [26].

For the primary endpoint analysis, the raw 24-h cough count was standardized to an average hourly count, then log-transformed as done previously [12, 27]. To detect a dose–response signal, four candidate dose–response models were tested with a single contrast test using the generalized MCP-Mod approach. The null hypothesis, “the response at all doses is equal,” was tested against the alternative, “there is a dose–response relationship.” If at least one of the four individual tests of models was statistically significant (adjusted p of one-sided test ≤ 0.1), a dose–response signal was considered established. The model with the best fit was then used for the estimation of the dose–response curve and the minimum effective dose (MED). For further information on the primary endpoint analysis, see the Supplementary Methods.

Sample size calculations were performed for establishing evidence of a drug effect across the doses. A sample size of 50 participants per dose group was predicted to have at least 85% power to demonstrate a dose–response relationship for the primary efficacy endpoint, using a one-sided test at a type I error rate of α = 0.10 (see the Supplementary Methods for more details).

The secondary endpoint analyses and definitions of the per protocol, full analysis and safety analysis sets are described in the Supplementary Methods.

Statistical evaluation was performed using SAS software version 9.4 or higher (SAS Institute, USA) or ValidR software version 3.5.2 or higher (Mango Solutions, UK). Confirmatory p-values are reported for the analysis of the primary endpoint. The study was not powered for individual pairwise comparisons between dose groups. Analysis of secondary endpoints, sensitivity analyses, and AEs should be interpreted as exploratory.

Results

Of 399 participants screened between October 2, 2020 and March 12, 2021, 310 were randomized to eliapixant 25 mg (n = 75), 75 mg (n = 78), 150 mg (n = 80), or placebo (n = 77) (Fig. 1). All randomized participants were included in the full and safety analysis sets. A total of 283 participants were included in the per protocol set (eliapixant 25 mg n = 67, 75 mg n = 69, 150 mg n = 73, placebo n = 74). In total, 276 participants (89%) completed the treatment period.

Fig. 1
figure 1

Participant disposition. Includes the 12-week treatment period and the 30-day safety follow-up. If a participant has more than one validity finding that excludes them from an analysis set, all the findings are displayed. All 34 participants (11%) who discontinued the treatment phase of the study entered the safety follow-up. A total of 294 participants (95%) completed the 30-day safety follow-up. COVID-19 corona virus disease 2019

Full size image

Baseline demographics and clinical characteristics were generally well balanced across the treatment groups (Table 1), although mean 24-h cough count (Table 2) and awake cough count in the eliapixant 150 mg group were slightly lower than in other groups. The baseline awake cough count was higher than the 24-h cough count in all treatment groups. Overall, 76 (27%) participants had a low baseline 24-h cough count of < 10 coughs per hour.

Table 1 Baseline demographics and clinical characteristics (per protocol set)
Full size table
Table 2 Change from baseline in 24-h cough count after 12 weeks of intervention (per protocol set)
Full size table

The data for the primary efficacy endpoint, change from baseline in 24-h cough count after 12 weeks of intervention, are shown in Table 2 and Fig. 2A. At Week 12, the 24-h cough count had decreased from baseline in all treatment groups. The largest relative and placebo-adjusted reductions in 24-h cough count from baseline were seen in the 75 mg eliapixant group.

Fig. 2
figure 2

(A) Change from baseline in 24-h cough count throughout study period and (B) the estimated dose–response Emax model for the change from baseline to Week 12 in log-transformed 24-h cough count with an 80% CI (per protocol set). In (B), circles indicate the estimated dose response in each dose group adjusted for baseline cough count and geographic region. The dotted horizontal reference line at − 0.44 represents the estimated dose response in the placebo group. The solid line indicates the estimated Emax dose–response model and the dashed lines indicate the 80% CI. CI confidence interval, Emax asymptotic maximum change from placebo effect

Full size image

For the primary analysis of the primary endpoint, a statistically significant dose–response signal was detected with eliapixant for change from baseline in 24-h cough count at Week 12, with multiplicity-adjusted p-values of < 0.1 for all four candidate models (Supplementary Table S1). As a result of the better model fit, the Emax model was used to derive the MED (Fig. 2B). The MED to achieve a relative change vs. placebo of − 20% (i.e., log[0.8] =  − 0.22 on the log-transformed scale (Fig. 2B)) was estimated at ~ 58 mg eliapixant twice daily.

Reductions in 24-h cough count with the two higher doses of eliapixant relative to placebo were observed early at Week 2, with further reductions at Week 4 and Week 8 (Fig. 2A). A ≥ 30% reduction from baseline in 24-h cough count at Week 12 was reported in 34 participants (46%) with placebo and 35–44 participants (52–64%) receiving eliapixant. Compared with placebo, more participants in the 75 mg group reached this responder threshold at Week 12 (mean treatment difference: 18%, 95% confidence interval [CI] 2 to 34, p = 0.03). A smaller treatment difference vs. placebo was observed for the other doses of eliapixant with a mean treatment difference of 6% (95% CI − 10 to 23, p = 0.5) and 8% (95% CI − 9 to 24, p = 0.4) for the 25 mg and 150 mg groups, respectively.

Similar findings to those for 24-h cough count were observed for the change from baseline in awake cough count at all study visits (see Supplementary Results and Supplementary Fig. S2 for more details).

Cough severity was reduced with all doses of eliapixant at Week 12 vs. baseline, with a small numeric reduction vs. placebo (Table 3). More participants in the 75 mg group experienced a ≥ 30-scale unit reduction in cough severity at Week 12 vs. placebo (mean treatment difference of 16%, 95% CI 1 to 31, p = 0.03). A smaller treatment difference vs. placebo was observed for the other doses of eliapixant (Table 3).

Table 3 Secondary efficacy cough-related endpoints relating to severity and QoL (per protocol set)
Full size table

There was a dose-dependent improvement in LCQ total score after 12 weeks. However, the differences vs. placebo were small (0.1, 95% CI − 1.0 to 1.2, in the 25 mg group; 0.4, 95% CI − 0.7 to 1.4, in the 75 mg group; 0.5, 95% CI − 0.6 to 1.6 in the 150 mg group) (Table 3). The percentage of participants with a ≥ 1.3-point increase in LCQ total score from baseline after 12 weeks was similar between all three doses of eliapixant and placebo.

Sensitivity analyses of the full analysis set confirmed the primary endpoint and secondary endpoint results in the per protocol set (data not shown).

AEs were reported in 39 participants (51%) with placebo and 43–51 participants (57–65%) receiving eliapixant, with most considered mild or moderate (Table 4). The proportion of participants reporting AEs (including those described as severe) was slightly higher in the two higher-dose eliapixant groups than the low-dose eliapixant or placebo groups (Table 4). The most frequently occurring AE was dysgeusia, which occurred in 1 participant (1%) in the placebo group and 1–13 participants (1–16%) in the eliapixant group in a dose-related manner (Table 5). Other AEs relating to taste or smell disorders were similarly more frequent with eliapixant than placebo (Table 5).

Table 4 Summary of treatment-emergent AEs (safety analysis set)
Full size table
Table 5 Most frequently reported treatment-emergent AEs, and AEs related to taste, bleeding, and drug-related hepatic disorders (safety analysis set)
Full size table

AEs leading to study drug discontinuation were more common with eliapixant than placebo (Table 4). An SAE of abnormal liver tests leading to study drug discontinuation occurred in 1 participant in the 150 mg eliapixant group and was reported as a suspected unexpected serious adverse reaction (SUSAR). No deaths occurred during the study.

Changes in some laboratory safety parameters were reported, including 2 participants receiving eliapixant (75 mg, n = 1; 150 mg, n = 1) who had alanine aminotransferase exceeding the three-fold upper limit of normal, which triggered close liver observation in accordance with the US Food and Drug Administration Guidance for Industry [28] and the study protocol. In the patient receiving eliapixant 150 mg, the SUSAR was considered a moderate drug-induced liver injury (DILI) of hepatocellular origin. The participant prematurely discontinued eliapixant at the 4-week visit because of the liver event, after which the liver enzyme values returned to normal. In the overall population, dose-dependent increases in mean and median values of alkaline phosphatase, fibrinogen, and plasma antithrombin III activity were observed. There were no relevant mean changes in other liver enzymes at any dose of eliapixant during treatment in the overall population. See the Supplementary Results for more details.

Thirty-one participants who spontaneously reported a taste-related AE during the treatment period completed an end-of-study assessment on taste disturbances (Fig. 3). The frequency and how bothersome the taste disturbances were in the eliapixant groups increased in a dose-related manner (Fig. 3A). No participants described the taste effects as “extremely” bothersome. An answer of “very” bothersome was only recorded in the 150 mg group (Fig. 3B).

Fig. 3
figure 3

End-of-study assessment on taste disturbances (safety analysis seta)

Full size image

Discussion

PAGANINI confirmed data from the phase 2a study suggesting that eliapixant is effective at reducing 24-h cough count in patients with RCC. The detection of a statistically significant dose–response signal with eliapixant was achieved for the primary endpoint of change from baseline in 24-h cough count at Week 12. At Week 12, 24-h cough count was reduced by 27% vs. placebo in the 75 mg group. In an analysis of secondary efficacy endpoints, awake cough count was also reduced by 28% with 75 mg eliapixant vs. placebo at Week 12. Compared with placebo, more participants in the 75 mg eliapixant group reached a ≥ 30% reduction in 24-h cough count and ≥ 30-scale unit reduction in cough severity at Week 12 from baseline; however, cough-related QoL as measured by LCQ total score did not improve.

The phase 2a study of eliapixant demonstrated similar reductions in 24-h cough and awake cough counts to those reported here [18]. However, the improvements in cough severity and cough-related QoL vs. placebo seen in the earlier study [18] were not observed to the same extent. While comparisons between studies should be made with caution, this observation may be explained by the larger placebo response seen in PAGANINI. However, improvements in cough severity and cough-related QoL were reported for the phase 2b gefapixant and sivopixant trials [12, 27], and the phase 3 COUGH-1 and COUGH-2 studies despite large placebo effects [13]. The lack of patient-perceived improvement in cough in this study is therefore difficult to explain. However, it should be noted that PAGANINI was not powered to detect significant differences in patient-reported outcome parameters between treatment groups.

In this study, the efficacy effects in the 150 mg eliapixant group were not greater than those in the 75 mg group, suggesting that a plateau in dose response was reached, as indicated by the estimated dose–response curve. This finding may have also been influenced by the less severe baseline cough characteristics in the 150 mg group. A plateau in dose response for reduction in cough count was also observed with eliapixant in the phase 2a study, although subjective endpoints continued to improve with the highest dose [18]. The plateau in dose response is also supported by data from healthy volunteers [16], whereby the two higher doses of eliapixant had similar trough plasma drug concentrations, and the plasma concentrations predicted to achieve ≥ 80% P2X3 receptor occupancy (the expected threshold for efficacy based on unpublished preclinical studies; data on file, Bayer AG) were reached with both higher doses [16]. Achievement of the primary endpoint and the low MED to achieve 20% improvement over placebo are therefore notable considering the globally heterogeneous study population, the high placebo response, the overall high number of participants experiencing a low baseline 24-h cough count of < 10 coughs per hour, and the lower baseline cough counts and efficacy results in the 150 mg eliapixant group.

The safety and tolerability profiles in PAGANINI are generally consistent with other studies of eliapixant in healthy subjects and the phase 2a study in patients with RCC [16,17,18]. However, a case of a moderate DILI of hepatocellular origin occurred during treatment with 150 mg eliapixant and contributed to the need for intensified liver monitoring in clinical trials with eliapixant. In a second participant, alanine aminotransferase levels exceeding the three-fold upper limit of normal led to close liver observation, and a dose-dependent increase in mean alkaline phosphatase levels in the overall population was observed during the treatment period. The clinical relevance of increased alkaline phosphatase levels is unclear, as is the origin (liver vs. bone) in the absence of a concurrent increase in the mean values of other liver enzymes. In the phase 2a study of the P2X3 antagonist sivopixant for RCC, a participant receiving sivopixant also experienced a DILI during the trial [29].

Taste-related AEs were reported in 24% of participants in the 150 mg eliapixant group with fewer reports in participants receiving lower doses. One participant discontinued treatment due to dysgeusia as part of a combination of nine AEs. No participants who spontaneously reported a taste-related AE described the effect as “extremely” bothersome. As with the phase 2b study of gefapixant [12], dysgeusia was the most reported AE in PAGANINI. However, taste-related AEs were previously reported in up to 81% of patients with gefapixant in phase 2b [12] compared with up to 24% of participants with eliapixant in this study. In phase 3 trials with gefapixant 45 mg, taste-related AEs were reported by 59% of participants at Week 12 in COUGH-1 and 69% of participants at Week 24 in COUGH-2 [13]. The smaller impact on taste perception with eliapixant may be due to its high selectivity for the P2X3 receptor leading to a low potential for off-target effects mediated by P2X2/3 receptors [16].

Strengths of PAGANINI included that the baseline demographics reflect those seen in the clinical RCC population [30]. Recruitment of participants across 19 countries means the results are likely to reflect the global population of patients with RCC. Limitations of the study include a lack of powered individual pairwise comparisons between dose groups; however, the aim of this study was to establish evidence of a drug effect across the doses to support the dose selection for phase 3 studies [25, 26].

In summary, the PAGANINI study showed that eliapixant was effective at reducing 24-h cough count vs. placebo in patients with RCC. The safety and tolerability profiles in PAGANINI were consistent with other studies of eliapixant in healthy subjects and the phase 2a study in patients with RCC. However, a case of a moderate DILI of hepatocellular origin occurred during treatment with 150 mg eliapixant. This DILI contributed to the need for intensified liver monitoring in clinical trials with eliapixant and the subsequent discontinuation of the entire development program in all indications by Bayer AG.