Efficacy and Tolerability of Gefapixant for Treatment of Refractory or Unexplained Chronic Cough

Key Points

Question

Is the use of gefapixant to treat refractory or unexplained chronic cough associated with greater benefits or harms compared with placebo?

Findings

In this dose-response meta-analysis that included 9 randomized clinical trials and 2980 patients, compared with placebo, use of gefapixant (45 mg orally twice daily) reduced cough frequency by 17.6%, decreased cough severity by 6.2 mm on the 100-mm visual analog scale, and improved cough quality of life by 1 point on the Leicester Cough Questionnaire (score range, 3-21). Patients treated with gefapixant (45 mg orally twice daily) had, however, a 32% increase in taste-related adverse events compared with placebo.

Meaning

Gefapixant may provide benefit to some patients with refractory or unexplained chronic cough, but the magnitude is likely small and the risk of adverse events, particularly taste-related, remains substantial.

Abstract

Importance

Gefapixant represents an emerging therapy for patients with refractory or unexplained chronic cough.

Objective

To evaluate the efficacy and tolerability of gefapixant for the treatment of adults with refractory or unexplained chronic cough.

Data Sources

MEDLINE, Embase, Cochrane Central Register of Controlled Trials, and Web of Science from November 2014 to July 2023.

Study Selection

Two reviewers independently screened for parallel and crossover randomized clinical trials (RCTs) that compared, in patients with refractory or unexplained chronic cough, either gefapixant with placebo, or 2 or more doses of gefapixant with or without placebo.

Data Extraction and Synthesis

Two reviewers independently extracted data. A frequentist random-effects dose-response meta-analysis or pairwise meta-analysis was used for each outcome. The GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) approach was used to rate the certainty in whether patients would perceive the effects as important (greater than the minimal important difference [MID]) or small (less than the MID).

Main Outcomes and Measures

Cough frequency (measured using the VitaloJAK cough monitor; MID, 20%), cough severity (measured using the 100-mm visual analog scale [VAS]; higher score is worse; MID, 30 mm), cough-specific quality of life (measured using the Leicester Cough Questionnaire [LCQ]; score range, 3 [maximal impairment] to 21 [no impairment]; MID, 1.3 points), treatment-related adverse events, adverse events leading to discontinuation, and taste-related adverse events.

Results

Nine RCTs including 2980 patients were included in the primary analysis. Compared with placebo, gefapixant (45 mg twice daily) had small effects on awake cough frequency (17.6% reduction [95% CI, 10.6%-24.0%], moderate certainty), cough severity on the 100-mm VAS (mean difference, −6.2 mm [95% CI, −4.1 to −8.4]; high certainty), and cough-specific quality of life on the LCQ (mean difference, 1.0 points [95% CI, 0.7-1.4]; moderate certainty). Compared with placebo, gefapixant (45 mg twice daily) probably caused an important increase in treatment-related adverse events (32 more per 100 patients [95% CI, 13-64 more], moderate certainty) and taste-related adverse events (32 more per 100 patients [95% CI, 22-46 more], high certainty). High-certainty evidence suggests that gefapixant (15 mg twice daily) had small effects on taste-related adverse events (6 more per 100 patients [95% CI, 5-8 more]).

Conclusions and Relevance

Compared with placebo, gefapixant (45 mg orally twice daily) led to modest improvements in cough frequency, cough severity, and cough-specific quality of life but increased taste-related adverse events.

This systematic review and dose-response meta-analysis evaluates the efficacy and tolerability of gefapixant, a new P2X3 antagonist, for the treatment of adults with refractory or unexplained chronic cough.

Introduction

Chronic cough affects 2% to 18% of adults globally¹ and can substantially affect quality of life.² Although some patients respond well to treatments targeting possible causes of cough, 10% to 60% have refractory or unexplained chronic cough.^3,4,5

With no licensed treatments for refractory or unexplained chronic cough, guidelines recommend treatment with speech pathology or neuromodulators.^6,7 Neuromodulators include morphine,⁸ gabapentin,⁹ pregabalin,¹⁰ and amitriptyline,¹¹ but these therapies have dose-limiting adverse effects, show poor response in some patients, and pose concerns about long-term use.⁶ Alternative therapies for refractory or unexplained chronic cough remain an unmet need.

A new class of antitussives for refractory or unexplained chronic cough includes P2X3 antagonists.^12,13,14 These agents bind to the P2X3 receptor, an adenosine triphosphate–gated ion channel on chemically sensitive airway C-fibers.^15,16,17 Blocking this receptor prevents adenosine triphosphate from opening the channel, thereby reducing activation of C-fibers that transmit action potentials to the brainstem and somatosensory cortex that control cough.

Clinical trials have highlighted the efficacy of P2X3 antagonism for treatment of refractory or unexplained chronic cough.^{12,13,14,18,19} Two phase 3 trials evaluating gefapixant met the primary end point of reducing 24-hour cough frequency compared with placebo.¹⁹ Given these trials, Japan and Switzerland have licensed gefapixant for treatment of refractory or unexplained chronic cough, while the US Food and Drug Administration (FDA), European Medicines Agency (EMA), Health Canada, and other major regulatory authorities deliberate. Although gefapixant is not yet approved by the FDA for use in the US, there remains a possibility that gefapixant could become the first licensed treatment for refractory or unexplained chronic cough in multiple countries globally.

Evidence of gefapixant’s magnitude of effect on outcomes of benefit and tolerability should precede its widespread use. No studies have previously evaluated all randomized trial evidence addressing this therapy—findings that could inform regulatory agencies, guideline developers, payers, clinicians, and patients. Therefore, a systematic review and dose-response meta-analysis was performed to evaluate the effects of gefapixant in patients with refractory or unexplained chronic cough.

Methods

This study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).²⁰ Prior to data analysis, we registered our protocol on PROSPERO (CRD42022383872). This article differs from the protocol in that it focuses on gefapixant rather than all P2X3 antagonists. Several reasons informed this protocol deviation: (1) of all P2X3 antagonists, gefapixant was most likely to become imminently available for use in clinical practice; (2) while we conducted the current study, drug development programs for eliapixant²¹ and sivopixant²² discontinued after their phase 2b trials; and (3) we believed that evidence syntheses on camlipixant were premature with phase 3 trials still ongoing.

Eligibility Criteria

We included parallel and crossover randomized clinical trials (RCTs) that compared gefapixant with placebo or alternative doses of the drug in adults (≥18 years) with refractory or unexplained chronic cough lasting longer than 8 weeks. We excluded trials in which less than 80% of patients had refractory or unexplained chronic cough lasting longer than 8 weeks.

Data Sources and Search

We searched MEDLINE, Embase, Cochrane Central Register of Controlled Trials, and Web of Science from November 2014 to July 2023 with no restrictions on language or publication status (eMethods 1 in Supplement 1). To ensure an exhaustive search for unpublished data, we hand-searched the FDA, EMA, and pharmaceutical company websites.

Study Selection

Paired reviewers independently screened titles, abstracts, and full-text articles. Title, abstract and full-text screening required agreement between reviewers. Reviewers resolved disagreements by discussion and, if necessary, by consultation with an adjudicator.

Data Extraction

Two reviewers (E.K., M.P.) extracted data independently using a standardized, pilot-tested form. eMethods 2 in Supplement 1 lists the study characteristics and baseline patient information captured from trials. Reviewers resolved discrepancies by discussion and, when necessary, by adjudication with a third reviewer.

Reviewers extracted (1) 24-hour, (2) awake, and (3) sleep cough frequencies, each measured using the VitaloJAK cough monitor (Vitalograph)²³; (4) cough severity, measured using the 100-mm visual analog scale (VAS)²⁴; (5) cough-specific quality of life, measured using the Leicester Cough Questionnaire (LCQ)²⁵ or Cough Quality-of-Life Questionnaire (CQLQ)²⁶; (6) treatment-related adverse events; (7) adverse events leading to discontinuation; and (8) taste-related adverse events. We captured all outcomes at the longest follow-up. For adverse events, we recorded the number of patients with 1 or more events.

Risk-of-Bias Assessments

Two independent reviewers (E.K., M.P.) used a modified Cochrane RoB 2.0 tool (Cochrane Bias Methods Group) for assessing risk of bias in RCTs.²⁷ Reviewers rated both parallel and crossover trials at “low risk of bias,” “probably low risk of bias,” “probably high risk of bias,” or “high risk of bias” across each domain. For crossover trials, we considered risk of bias arising from period and carryover effects.²⁷ Reviewers resolved discrepancies by discussion and, when necessary, by adjudication with a methodologist.

Data Synthesis

We conducted random-effects dose-response meta-analysis with the restricted maximum likelihood estimator.^28,29 Dose-response meta-analysis summarizes, across studies, the relationship between treatment dose and an outcome. Nonlinearity was modeled using 1-stage restricted cubic splines,³⁰ with knots set at 10%, 50%, and 90%. Wald-type tests tested nonlinearity, and goodness-of-fit statistics informed model fit.

For outcomes in which we found no evidence of a dose-response relationship, we assumed the same effect across doses and performed a random-effects pairwise meta-analysis.

For cough frequency outcomes, we calculated pooled mean differences in log-transformed coughs/h, with 95% CIs. For studies in which only means and standard deviations of the unlogged data were available, we approximated means and standard deviations on the log scale.³¹ When reported, mean changes in log-transformed coughs/h from baseline estimated using analysis of covariance or linear mixed-effects models that adjusted for baseline covariates were preferentially recorded. To calculate percent reductions, we exponentiated mean differences in log-transformed coughs/h using 100*(e^diff – 1).

For patient-reported outcomes, we calculated pooled mean differences with 95% CIs. For cough-specific quality of life, if a trial reported the outcome using the CQLQ instead of the LCQ, we assumed both instruments measured the same construct and transformed mean differences on the CQLQ into natural units of the more widely used LCQ.³²

For all efficacy outcomes apart from sleep cough frequency, we modeled risk differences and 95% CIs for achieving at least the minimal important difference (MID). The MID is the smallest change in an outcome that patients would perceive as important. Based on prior work, we considered MIDs as a 20% reduction in cough frequency,³³ a 30-mm reduction on the 100-mm cough severity VAS,³⁴ and a 1.3-point increase on the LCQ.³⁵ Given that regulators would most likely approve a dose of 45 mg twice daily, we focused our modeling of risk differences for achieving the MID based on summary mean differences comparing the effects of gefapixant at this dose with placebo.³⁶

We calculated pooled relative risks (RRs) with 95% CIs for adverse event outcomes. We also determined absolute risk increases using estimates of baseline risk from the median of the control group from eligible trials.

For crossover trials, we prioritized analysis of both periods over an analysis of the first period only. We pooled crossover trials with parallel designs using methods outlined in the Cochrane Handbook.³⁷

In the primary analyses, we included only published trials. To test the impact of including unpublished trials, we ran, for applicable outcomes, post hoc sensitivity analyses including unpublished trials.

Analyses were among the intention-to-treat population with complete case data. Although we included trials that evaluate any dose of gefapixant, to avoid spurious judgments based on imprecise models, we limited our inferences to clinically relevant doses and those for which the model yielded credible results (0-60 mg twice daily). We ran analyses using the dosresmeta and metafor packages in R version 4.2.1 (R Foundation for Statistical Computing).

Heterogeneity and Meta-Regression

For 2-stage linear dose-response models, we examined, across studies, heterogeneity in the study-specific dose-response association through a combination of visual inspection of dose-response plots, χ² tests, and the I² statistic. For 1-stage dose-response models using restricted cubic splines, we examined heterogeneity through visual inspection of the dose-response plots and the variance partition coefficient (VPC).³⁰ The VPC is a study-specific and dose-specific statistic that represents the percentage of heterogeneity out of the total variability specific to the study. For nonreferent doses in each study, we calculated a VPC and interpreted the average of the study-specific VPCs by dose as the dose-specific I².³⁰

To explain heterogeneity, we performed meta-regression to test the following a priori hypotheses about study-level covariates: (1) crossover trials show larger effects than parallel trials; and (2) trials with higher risk of bias show larger effects than trials with lower risk of bias. We performed the last meta-regression only if we found considerable variability within a risk-of-bias component. Tests of interaction established whether treatment effects differed significantly.

Rating the Certainty of Evidence

Using the GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) approach, 2 reviewers (E.K., I.S.) assessed the certainty in evidence for all doses and outcomes.³⁸ These reviewers rated the certainty in whether patients would perceive effects to be important (ie, greater than the MID) or small (less than the MID). To ascertain MIDs for adverse event outcomes, we surveyed 5 patient partners with refractory or unexplained chronic cough. They informed absolute MIDs of 15% for treatment-related adverse events, 10% for adverse events leading to discontinuation, and 10% for taste-related adverse events.

Results

After screening 713 titles and abstracts and 100 full-text articles, we identified 11 RCTs; of these, 2 were unpublished and only reported results on the clinical trial registration (Figure 1). For the primary analysis, we included 9 published RCTs including 2980 patients; of these 9 trials, 4 had a crossover design^12,39,40 and 5 had a parallel design.^19,41,42 All trials, apart from 1 that addressed alternative doses of gefapixant,⁴² compared 1 or more doses with a placebo. Industry funded all trials. Trials excluded people who smoke; enrolled patients had a mean age of 58.5 years, a median cough duration of 11.6 years, and a mean of 75.7% were female—characteristics typical for those with refractory or unexplained chronic cough. The Table lists the characteristics of the 9 included studies.

Figure 1. Literature Search Flow Diagram: Study of Gefapixant for Chronic Cough.

EMA indicates European Medicines Agency; FDA, US Food and Drug Administration.

Table. Characteristics of Included Studies.

Source	Design, country	Participants, No.	Age, mean (SD), y	Female, %	Duration of cough, mean (SD), y	Eligibility threshold	Intervention	Outcomes
Phase 3
COUGH-2 McGarvey et al,19 2022 (NCT03449147)	Parallel RCT Multinational	1314	58.1 (12.1)	74.9	11.2 (9.8)	Cough severity VAS ≥40 mma	Gefapixant (15 mg orally twice daily for 52 wk); gefapixant (45 mg orally twice daily for 52 wk); placebo	24-h Cough frequency, awake cough frequency; cough severity; cough-specific quality of life; treatment-related adverse events; adverse events leading to discontinuation; taste-related adverse events
COUGH-1 McGarvey et al,19 2022 (NCT03449134)	Parallel RCT Multinational	730	59.0 (12.6)	74.2	11.6 (9.5)	Cough severity VAS ≥40 mma	Gefapixant (15 mg orally twice daily for 52 wk); gefapixant (45 mg orally twice daily for 52 wk); placebo	24-h Cough frequency, awake cough frequency; cough severity; cough-specific quality of life; treatment-related adverse events; adverse events leading to discontinuation; taste-related adverse events
Niimi et al,42 2022 (NCT03696108)	Parallel RCT Japan	169	57.9 (15.0)	62.7	9.2 (10.5)	NR	Gefapixant (15 mg orally twice daily for 52 wk); gefapixant (45 mg twice daily for 52 wk)	Cough-specific quality of life; treatment-related adverse events; adverse events leading to discontinuation; taste-related adverse events
Phase 3b
McGarvey et al,43 2023 (NCT04193202)	Parallel RCT Multinational	419	52.6 (13.7)	64.7	0.6 (0.2)	Cough duration >8 wk for <12 mo	Gefapixant (45 mg orally twice daily for 12 wk); placebo	Cough severity; cough-specific quality of life; adverse events leading to discontinuation; taste-related adverse events
Phase 2
Smith et al,40 2020 (NCT02349425)	Crossover RCT US	30	60.2 (11.1)	80.0	13.2 (1.9-42.8)b	Cough severity VAS ≥40 mma	Gefapixant (7.5/15/30/50 mg orally twice daily, each for 4 d); placebo	24-h Cough frequency, awake cough frequency; sleep cough frequency; cough severity; cough-specific quality of life; treatment-related adverse events; adverse events leading to discontinuation; taste-related adverse events
Smith et al,40 2020 (NCT02349425)	Crossover RCT US	29	63.2 (7.4)	86.2	15.4 (1.4-55.3)b	Cough severity VAS ≥40 mma	Gefapixant (50/100/150/200 mg orally twice daily, each for 4 d); placebo	24-h Cough frequency, awake cough frequency; sleep cough frequency; cough severity; cough-specific quality of life; treatment-related adverse events; adverse events leading to discontinuation; taste-related adverse events
Abdulqawi et al,12 2015 (NCT01432730)	Crossover RCT UK	24	54.5 (11.1)	75.0	9.0 (3-25)c	NR	Gefapixant (600 mg orally twice daily for 2 wk); placebo	24-h Cough frequency; awake cough frequency; sleep cough frequency; cough severity; cough-specific quality of life; adverse events leading to discontinuation; taste-related adverse events
Morice et al,39 2019 (NCT02476890)	Crossover RCT UK	12	61.1 (8.7)	87.5	14.6 (9.9)	HARQ score >20	Gefapixant (50 mg orally twice daily for 1 d); placebo	24-h Cough frequency; cough severity; adverse events leading to discontinuation; taste-related adverse events
Phase 2b
Smith et al,41 2020 (NCT02612610)	Parallel RCT UK, US	253	60.2 (9.9)	76.3	14.5 (11.7)	Cough severity VAS ≥40 mma	Gefapixant (7.5 mg orally twice daily for 12 wk); gefapixant (20 mg orally twice daily for 12 wk); gefapixant (50 mg orally twice daily for 12 wk); placebo	24-h Cough frequency, awake cough frequency; sleep cough frequency; cough severity; cough-specific quality of life; treatment-related adverse events; adverse events leading to discontinuation; taste-related adverse events

Abbreviations: HARQ, Hull Airway Reflux Questionnaire; NR, not reported; RCT, randomized clinical trial; VAS, visual analog scale.

^a Range, 0-100 mm; higher score is worse.

^b Reported only median (range).

^c Reported only median (IQR).

Risk of Bias in Included Studies

Among the 9 included RCTs, reviewers rated 2 at risk of bias for at least 1 domain (eTable 1 in Supplement 1). One trial did not adequately describe concealed allocation, and 1 trial had 20% or more missing outcome data that could lead to bias.

Outcome Data

eTable 2 in Supplement 1 presents evidence profiles for the effects of gefapixant at twice-daily doses of 15, 30, 45, and 60 mg. eTable 3 in Supplement 1 reports the goodness-of-fit statistics for all dose-response models.

24-Hour Cough Frequency

Six RCTs including 2472 patients reported 24-hour cough frequency measured by the VitaloJAK ambulatory cough recorder.^12,19,40,41 We did not find evidence of a nonlinear relationship between increasing gefapixant dose and reduction in 24-hour cough frequency (Figure 2A). The linear model demonstrated that, compared with placebo, gefapixant (45 mg twice daily) probably had a small effect on 24-hour cough frequency (16.0% reduction [95% CI, 9.4%-22.0%]; risk difference for achieving at least the MID, 4.7% [95% CI, 0.6%-8.7%]; moderate certainty) (Figure 3; eTable 4 in Supplement 1), whereas at 60 mg twice daily, gefapixant probably caused an important reduction in 24-hour cough frequency (20.7% reduction [95% CI, 12.5%-28.3%]; moderate certainty).

Figure 2. Dose-Response Relationship Between Gefapixant Dose and 24-Hour, Awake, and Sleep Cough Frequency.

A and B, Percent reductions are in comparison with placebo (dose = 0). Circles represent percent reductions observed in individual trials, and the diameters are proportional to the inverse variance of the study. Randomized clinical trials (RCTs) evaluating different doses may have contributed more than 1 circle in the graph. Doses are truncated at the largest dose for use in clinical practice (60 mg twice daily). C, Effects for sleep cough frequency are represented as mean differences in log-transformed coughs/h. Percent reduction compared with placebo is calculated as 100*(e^diff – 1), where diff is the mean difference in log-transformed coughs/h. Size of data markers indicates the weight of each study in the analysis.

Figure 3. Effects of Gefapixant by Approximate Dose Compared With Placebo Estimated by a Study-Level Meta-Analysis.

Model estimates are listed at clinically relevant doses from 0 to 60 mg twice daily at equal intervals (every 15 mg twice daily). Cells in the last 4 columns show effect estimates and 95% CIs in parentheses. Shading corresponds to the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) certainty rating in whether patients would perceive effects to be important (ie, greater than the minimal important difference [MID]) or not importantly different (less than the MID). MIDs were considered as a 20% reduction in cough frequency, a 30-mm reduction on the 100-mm cough severity visual analog scale (VAS), a 1.3-point increase on the Leicester Cough Questionnaire (LCQ), a 15% increase for treatment-related adverse events, a 10% increase for adverse events leading to discontinuation, and a 10% increase for taste-related adverse events. Outcomes are ordered by the frequency with which studies consider the outcome as a primary, secondary, or safety outcome. eTable 2 in Supplement 1 reports the GRADE evidence profiles for each outcome and dose.

^aExpected risk or change from baseline with placebo.

^bNo evidence was found of a dose-response relationship for sleep cough frequency. Estimate is based on a pairwise meta-analysis treating all doses as having the same effect.

^cHigher score is worse; score range, 0 (no cough) to 100 mm (worst possible cough).

^dHigher score is better; score range, 3 (maximal impairment) to 21 (no impairment).

^eTaste-related adverse events include ageusia, dysgeusia, and/or hypogeusia.

Awake Cough Frequency

Six RCTs including 2452 patients reported awake cough frequency measured by the VitaloJAK ambulatory cough recorder.^12,19,40,41 Although we found evidence of a nonlinear relationship between dose and reduction in awake cough frequency, we found heterogeneity explained by a significant difference between parallel and crossover trials. Since crossover trials demonstrated larger effects than parallel trials, we modeled the dose-response relationship only among the parallel trials (3 RCTs, 2145 patients).^19,41 The dose-response model demonstrated that, compared with placebo, gefapixant (45 mg twice daily) probably had a small effect on awake cough frequency (17.6% reduction [95% CI, 10.6%-23.7%]; risk difference for achieving at least the MID, 6.9% [95% CI, 3.3%-10.5%]; moderate certainty) (Figure 3 and Figure 2B; eTable 4 in Supplement 1). At 60 mg twice daily, gefapixant probably caused an important reduction in awake cough frequency (22.8% reduction [95% CI, 14.0%-30.7%]; moderate certainty). A sensitivity analysis including 1 unpublished RCT⁴⁴ found similar results (eTable 5 in Supplement 1).

Sleep Cough Frequency

Four RCTs including 380 patients reported sleep cough frequency measured by the VitaloJAK ambulatory cough recorder.^12,40,41 We found no evidence of a relationship between gefapixant dose and reduction in sleep cough frequency. In a pairwise meta-analysis treating all doses as having the same effect, we found low-certainty evidence that gefapixant reduced sleep cough frequency compared with placebo (13.1% [95% CI, 47.8% reduction to 44.8% increase]; low certainty) (Figure 3 and Figure 2C).

Cough Severity

Eight RCTs including 2666 patients reported cough severity using the VAS.^{12,19,39,40,41,43} We found evidence of a nonlinear relationship between gefapixant dose and reduction in cough severity, but parallel and crossover trials significantly differed. Since crossover trials demonstrated larger effects than parallel trials, we modeled the dose-response relationship only among parallel trials (4 RCTs, 2292 patients).^19,41,43 The dose-response model demonstrated that, compared with placebo, gefapixant (45 mg twice daily) had a small effect on cough severity measured using the 100-mm VAS (mean difference, −6.2 mm [95% CI, −4.1 to −8.4]; risk difference for achieving at least the MID, 11.0% [95% CI, 6.8%-15.2%]; high certainty) (Figure 3 and Figure 4A; eTable 4 in Supplement 1).

Figure 4. Dose-Response Relationship Between Gefapixant Dose and Cough Severity, Cough-Specific Quality of Life, Taste-Related Adverse Events, and Adverse Events Leading to Gefapixant Discontinuation.

Mean differences or relative risks are in comparison with placebo (dose = 0). Shaded regions represent 95% CIs. Circles represent mean differences or relative risks observed in individual trials, and the diameters are proportional to the inverse variance of the study. Trials evaluating different doses may have contributed more than 1 circle in the graph. Doses are truncated at the largest dose for use in clinical practice (60 mg twice daily). For panels B through D, outcomes were analyzed using a 1-stage dose-response model using restricted cubic splines. Heterogeneity was examined through visual inspection of the dose-response plots and the variance partition coefficient (eFigures 1, 3, and 4 in Supplement 1). A, Assessed using visual analog scale (VAS). Higher score is worse; score range, 0 (no cough) to 100 mm (worst possible cough); minimum important difference (MID), 30 mm. B, Assessed using the Leicester Cough Questionnaire (LCQ). Higher score is better; score range, 3 (maximal impairment) to 21 (no impairment); MID, 1.3 points. C, Taste-related adverse events include ageusia, dysgeusia, and/or hypogeusia.

Cough-Specific Quality of Life

Eight RCTs including 2651 patients reported cough-specific quality of life.^{12,19,40,41,42,43} The dose-response model suggested that, compared with placebo, use of gefapixant (45 mg twice daily) resulted in a mean increase of 1 point (95% CI, 0.7-1.4) on the LCQ measuring cough-specific quality of life (risk difference for achieving at least the MID, 9.2% [95% CI, 5.6%-12.8%]; moderate certainty) (Figure 3 and Figure 4B; eTable 4 in Supplement 1).

Treatment-Related Adverse Events

Five RCTs including 2580 patients reported treatment-related adverse events.^19,40,41 The dose-response model demonstrated that, compared with placebo, use of gefapixant (45 mg twice daily) resulted in more treatment-related adverse events (RR, 2.7 [95% CI, 1.7-4.4]; moderate certainty). Treatment-related adverse events were lower with gefapixant (15 mg twice daily) but were still increased compared with placebo (RR, 1.4 [95% CI, 1.2-1.6]; moderate certainty) (Figure 3; eFigure 2 in Supplement 1).

Taste-Related Adverse Events

Nine RCTs including 2974 patients reported taste-related adverse events.^{12,19,39,40,41,42,43} The dose-response model suggested that, compared with placebo, gefapixant (45 mg twice daily) caused a significant increase in taste-related adverse events (RR, 9.0 [95% CI, 6.5-12.6]; high certainty), whereas at 15 mg twice daily, gefapixant caused a smaller increase (RR, 2.5 [95% CI, 2.2-2.9]; high certainty) (Figure 3 and Figure 4C). To explore the impact of including 2 unpublished RCTs,^44,45 we included these trials in a sensitivity analysis, which found similar results (eTable 5 in Supplement 1).

Adverse Events Leading to Discontinuation

Nine RCTs including 2974 patients reported adverse events leading to discontinuation.^{12,19,39,40,41,43} Modeling of the dose-response relationship found significant nonlinearity between gefapixant dose and adverse events leading to discontinuation. The nonlinear model suggested that, compared with placebo, gefapixant (45 mg twice daily) caused an important increase in adverse events leading to discontinuation (RR, 4.3 [95% CI, 3.2-5.8]; moderate certainty) and that gefapixant (15 mg twice daily) caused a small increase (RR, 1.9 [95% CI, 1.6-2.2]; high certainty) (Figure 3 and Figure 4D). A sensitivity analysis including 2 unpublished RCTs^44,45 found similar results (eTable 5 in Supplement 1).

Discussion

Compared with placebo, gefapixant (45 mg twice daily) resulted in improvements in cough frequency, severity, and quality of life that were less than the MID thresholds for patient importance. Gefapixant at this dose increased both treatment-related and taste-related adverse events by an absolute 32%—frequencies greater than the thresholds the patient partners considered important. At 60 mg twice daily, gefapixant caused significant reductions in cough frequency, but high-certainty evidence suggests that this dose not only increased treatment-related and taste-related adverse events but also increased adverse events that led to discontinuation of gefapixant.

Because reporting of average effects rests on an assumption that all patients experience similar benefit, the risk differences were modeled for achieving at least the MID for gefapixant at 45 mg twice daily. For awake cough frequency, compared with placebo, only 6.9% more patients receiving gefapixant would achieve a 20% change from baseline. However, greater risk differences were found for achieving at least a 30-mm reduction on the 100-mm cough severity VAS (11.0%) and a 1.3-point increase on the LCQ (9.2%). This modeling has statistical flaws including reduced power for statistical tests and failure to discern variation in treatment effects from random error⁴⁶; nevertheless, some may consider these proportions benefiting.

Although this study focused on the effects of gefapixant compared with placebo, substantial improvements in cough were found with administration of placebo in these studies. Among patients who received placebo, awake cough frequency improved by 54.8%, cough severity decreased by 24.2 mm on the VAS, and cough-specific quality of life increased by 3.0 points on the LCQ. The reasons for such large placebo responses, and whether placebo effects occur in clinical practice, remain unclear. Researchers have speculated that responses to placebo might occur due to changes in expectation among trial participants; investigator enthusiasm influencing patient expectation; natural disease progression; regression to the mean; activation of the Hawthorne effect among trial participants who receive regular observation⁴⁷; and/or reactivation of descending inhibitory control neurons by release of endogenous opioids.^48,49,50 Future studies should provide better understanding of the reasons for such large placebo responses.

The results of this study align with those of the largest trials¹⁹ evaluating gefapixant for refractory or unexplained chronic cough. At 12 weeks, the COUGH-1 trial reported that gefapixant (45 mg twice daily) was associated with a 18.5% reduction (95% CI, 0.9%-32.9%) in 24-hour cough frequency compared with placebo, while the COUGH-2 trial reported that, at 24 weeks, gefapixant (45 mg twice daily) was associated with a 14.6% reduction (95% CI, 1.4%-26.1%).¹⁹ Similar effect sizes were found, but with greater precision, for gefapixant at this dose (16.0% reduction [95% CI, 9.4%-22.0%]). Findings of a dose-dependent increase in the risk for adverse events, particularly taste-related, relate to studies that document unselective blocking of P2X3 and P2X2/3 receptors by gefapixant.^12,19 Inhibition of P2X2/3 heterotrimeric receptors on taste buds may lead to taste disturbances including ageusia, hypogeusia, and dysgeusia. One recent study provides reassuring evidence, however, that taste-related adverse events resolve, with 25% of patients having resolution of all taste adverse events while taking gefapixant, 63% experiencing resolution after stopping therapy, and 8% having resolution of 1 or more taste-related adverse events before and after treatment.⁵¹

Implications for Practice and Research

Future clinical trials including both placebo and untreated controls should discern what proportion of the placebo response in these trials relates to true placebo effects and/or nonspecific factors (ie, natural course of disease, regression to the mean).

Because current therapies for refractory or unexplained chronic cough pose substantial harms and have only low-quality evidence to support their benefits, gefapixant, despite its limited effectiveness, may still prove superior to these treatments. If such is the case, gefapixant may provide an option for patients who have tried and failed all other therapies for identifiable causes. The small benefits of gefapixant and limitations of existing therapies underscore the urgent need for research to identify interventions with greater impact.

Strengths

This study has several strengths. By treating dose as a continuous variable, the study avoided misleading aggregation of treatment effects across studies that evaluated different doses. A comprehensive search of databases, regulatory sites, and conference proceedings mitigated the risk of publication bias. To further reduce bias, 2 reviewers independently screened studies, abstracted data, and performed risk-of-bias assessments. Using the GRADE approach, the certainty in whether patients would consider the effects of gefapixant across doses and outcomes as important was highlighted, thus providing possible insights regarding how patients are likely to experience the trade-off between its benefits and harms.

Limitations

This study has several limitations. First, it is limited by the dose range for which credible inferences can be drawn. Due to fewer trials examining doses beyond 60 mg twice daily, the study’s models yielded less precise results at higher doses. Second, studies excluded people who smoke or those who did with more than a 20 pack-year history. It remains unclear whether gefapixant has any role in the treatment for people who smoke or those with chronic obstructive pulmonary disease. Third, although trials recruited patients who had chronic cough refractory to at least 2 months of stable guideline-suggested therapy, lack of reporting regarding objective confirmation of diagnosis (ie, asthma reversibility, bronchial challenges), assessment of compliance with prior treatment trials, and/or independent adjudication that investigators followed the American College of Chest Physicians⁷ or European Respiratory Society⁶ guidelines reduced certainty about whether patients received adequate workup before entering the trials. Fourth, the modeling of risk differences for achieving at least the MID assumed similar standard deviations of the outcome measurements in the treatment and control groups, and that change from baseline scores in both groups are normally distributed. There remains a possibility that these assumptions were not met. Fifth, responder analyses have statistical limitations, a crucial one being that they fail to discern variation in treatment effects from random error.⁴⁶ Strong inferences about patients who potentially benefit from gefapixant solely based on the responder analyses are cautioned against.

Conclusions

Compared with placebo, gefapixant (45 mg orally twice daily) led to modest improvements in cough frequency, cough severity, and cough-specific quality of life but increased taste-related adverse events.

Supplement 1.

eMethods 1. Search Strategy

eMethods 2. Study Characteristics and Baseline Patient Information Collected From Trials

eMethods 3. PRISMA Checklist

eTable 1. Risk of Bias in Included Studies

eTable 2. GRADE Evidence Profile

eTable 3. Goodness-of-Fit Statistics for All Dose-Response Models

eTable 4. Risk Differences for Achieving a Benefit Greater Than or Equal to the MID for Gefapixant (45 mg Twice Daily)

eTable 5. Results of the Sensitivity Analyses Excluding Unpublished Trials

eFigure 1. Plot of Variance Partition Coefficient for the Meta-Analysis Between Gefapixant Dose and Cough-Specific Quality of Life

eFigure 2. Dose-Response Model Between Gefapixant Dose and Treatment-Related Adverse Events

eFigure 3. Plot of Variance Partition Coefficient for the Meta-Analysis Between Gefapixant Dose and Adverse Events Leading to Discontinuation

eFigure 4. Plot of Variance Partition Coefficient for the Meta-Analysis Between Gefapixant Dose and Taste-Related Adverse Events

Supplement 2.

Data Sharing Statement