Article Text

Patient preferences for the treatment of chronic cough: a discrete choice experiment
  1. Aparna C Swaminathan1,2,
  2. Jui-Chen Yang1,
  3. Helen Ding3,
  4. Kiran Grover4,
  5. Theresa Coles5,
  6. Jonathan Schelfhout3 and
  7. F Reed Johnson1,5
  1. 1Duke Clinical Research Institute, Durham, North Carolina, USA
  2. 2Duke University Medical Center, Durham, North Carolina, USA
  3. 3Merck & Co Inc, Rahway, New Jersey, USA
  4. 4Icahn School of Medicine, New York, New York, USA
  5. 5Duke University, Durham, North Carolina, USA
  1. Correspondence to Aparna C Swaminathan; aparna.swaminathan{at}


Background Chronic cough is common, negatively affects quality of life and has limited treatment options. Inhibition of purinergic signalling is a promising therapeutic approach but is associated with taste-related adverse effects. Little is known about treatment preferences from the perspective of patients with chronic cough, such as trade-offs between efficacy and side effect.

Methods Patients with chronic cough completed an online discrete choice experiment survey in which they answered a series of questions requiring a choice between two constructed treatment options characterised by varying attribute levels. Selection of cough and taste-related attributes was informed by qualitative interviews and clinical trial data. Logit-based models were used to analyse resulting choice data.

Results The discrete choice experiment survey was completed by 472 participants with chronic cough. Among study attributes, frequency of intense cough attacks was the most important to participants, followed by taste change, frequency of night-time coughing and frequency of daytime coughing. To accept the least preferred taste disturbance of a bitter, metallic, chalky or oily taste change, participants required either: (1) elimination of night-time cough along with a slight reduction in daytime cough; (2) elimination of daytime cough along with a pronounced reduction in night-time or (3) reduction in intense cough attacks from 7 to 2 times per week. Two distinct preference patterns were identified, each placing different importance on efficacy versus side effect trade-offs.

Conclusions Participants with chronic cough were willing to accept some taste disturbances in exchange for improved efficacy of chronic cough treatments. Knowledge of patient preferences can facilitate shared decision-making.

  • Cough/Mechanisms/Pharmacology

Data availability statement

Deidentified patient data are available on reasonable request.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Chronic cough is a common health problem that negatively affects quality of life. Emerging treatments for chronic cough are associated with taste disturbances, but the extent to which patients are willing to tolerate such side effects in exchange for cough relief is unknown.


  • Adults with chronic cough are generally willing to accept taste disturbances in exchange for improvement in chronic cough symptoms, particularly if the frequency of intense cough attacks is reduced.


  • Knowledge of the relative importance of cough-related attributes and potential treatment side effects from a patient perspective can support decision-making from the clinical and regulatory perspective.

Chronic cough is a common health problem that affects 5%–10% of adults worldwide and negatively impacts quality of life.1 Defined in adults as a cough that persists 8 weeks or longer, chronic cough is associated with sleep disturbance, fatigue, decreased work producitivty, anxiety, depression and increased healthcare utilisation.2 Chronic cough is commonly associated with comorbidities such as asthma, postnasal drip or gastro-oesophageal reflux disease, and in some patients, treatment of the comorbid condition can improve the cough.3 However, a considerable proportion of individuals have persistent cough despite a thorough evaluation and optimal treatment of comorbid conditions according to best practice guidelines, termed refractory or unexplained chronic cough.4 Recent insights into mechanisms of chronic cough has led to the development of drugs targeting neuronal hypersensitivity.5 Specifically, inhibition of afferent nerve excitability through ATP-dependent P2X3 receptor blockade has been identiifed as a promising therapeutic approach.6

Recent phase 3 clinical trials COUGH-1 and COUGH-2 evaluated P2X3 receptor antagonist gefapixant in adults with refractory chronic cough or unexplained chronic cough and demonstrated significant reduction in 24-hour cough frequency and improvement in cough-specific quality of life in patients treated with gefapixant 45 mg two times per day.7 8 However, taste-related adverse events were common, occurring in about 59%–69% of patients receiving gefapixant 45 mg two times per day.7 The most common taste-related adverse event was dysgeusia, often described as a metallic, bitter or salty taste. Other taste-related adverse events included ageusia (loss of taste sensitivity), hypogeusia (defined as diminished taste) and hypergeusia (increased taste sensitivity). Notably, unpleasant taste-related side effects have been observed in most P2X3 receptor antagonists in development.9 Despite the possible trade-offs between cough suppression and taste-related side effects, the extent to which patients are willing to tolerate such side effects in exchange for improvement in cough, and which cough-related aspects would be prioritised, is unknown.

Discrete choice experiments (DCEs) can quantify patients’ health preferences and their willingness to accept trade-offs among different health and treatment-related attributes. DCE respondents complete several choice questions in which they choose between treatment profiles which involve trade-offs among outcomes such as safety, efficiacy and convenience. Quantification of patient preferences is an effective expression of the patients’ voice, which has been prioritised by both clinical and regulatory agencies.10

This study leveraged a DCE in adults with chronic cough to understand preferences for outcomes relevant to chronic cough treatments. Selection of cough-related outcomes was informed by qualtitative interviews with adults with chronic cough in order to identify meaningful attributes that impacted preferences. The DCE survey measured the relative importance of potential treatment benefits and side effects and quantified patients’ willingness to accept risks of taste alteration side effects in exchange for improvements in chronic cough symptoms. An exploratory analysis evaluated how preferences vary across patients with chronic cough.


DCE development and patient involvement

DCE development was informed by the literature review of both chronic cough and taste disturbance, physician input and one-on-one qualitative interviews conducted with 20 adults diagnosed with chronic cough according to the American College of Chest Physicians’ guidelines.11 12 The literature review specifically focused on how adults with chronic cough define cough severity, using attributes such as frequency, intensity and timing of cough, as well as the impact of chronic cough on daily activities. With regard to taste disturbance, the literature review focused on how adults with dysgeusia or ageusia described their experiences. Taste-related attributes also were informed by recent clinical trial data (NCT02612610). Physicians involved in the care of patients with chronic cough provided input regarding aspects of chronic cough symptoms, diagnosis and management that were used to inform the design of qualitative interviews. The one-on-one qualitative interviews elicited patients’ experiences with chronic cough to identify chronic cough characteristics or attributes important to patients (Coles et al13 in press).14 The interviews also included bidding games and a card-sort activity to assess the importance of key chronic cough and treatment features. The final DCE attributes and levels were thus based directly on patient involvement.

Table 1 describes the final DCE attributes and levels. Cough-related attributes included cough frequency (0, 5, 10 or 20 times per hour), cough timing (daytime vs night-time) and number of intense cough attacks (0, 1, 3 or 7 times per week). Intense cough attacks were defined as a series of repeated violent coughs that last between 30 s and 2 min and can make it hard to breathe. Taste alteration was selected as the primary treatment side effect, with the following attribute levels: normal taste, inability to taste sweet foods, inability to taste salty foods, or a metallic, chalky or oily taste.

Table 1

Discrete choice experiment survey attributes and levels

To obtain trade-off data, the survey instructed respondents to assume they were coughing 20 times per hour both during the night and during the day and having 7 intense cough attacks per week. This represented the reference condition, in which respondents were not experiencing any taste changes. The DCE survey consisted of 10 versions of 12 choice questions that were generated using the SAS system15 to produce required statistical information to identify all attribute levels. Respondents were assigned to 1 of the 10 versions randomly. In each choice question, respondents were asked to choose between two constructed treatment options characterised by varying attribute levels from the reference condition. An example choice question is shown in figure 1.

Figure 1

Example choice question.

DCE respondents’ cough attributes at the time when his/her cough was at its worst were assessed using a five-level categorical scale. The questions were derived from the Cough Severity Diary, a patient-reported outcome measure with validity evidence in chronic cough.16 To test the draft DCE survey instrument, pretest interviews were conducted with patients with chronic cough, which are further detailed in online supplemental file 1. Several data quality checks were included in this study, which are detailed in online supplemental file 1.


For qualitative interviews and survey pretest interviews, adults diagnosed with chronic cough at Duke University who could read, speak and understand English were recruited. Purposive sampling for qualitative interviews targeted participants with diverse experiences with regard to time since chronic cough diagnosis, gender, age and race. For the DCE survey, adults with chronic cough were recruited through the Lightspeed Health online panel. Specifically, adults (≥18 years) were considered eligible if they had experienced a chronic cough of 8 weeks or longer in duration. Participants were also asked if they had a physician’s diagnosis of chronic cough. Purposeful sampling for the DCE survey targeted diversity with regard to age, race and education.

Statistical analyses

Preference weights

A random-parameter logit (RPL) model in Stata was used to estimate relative preference weights for all attribute levels and trade-offs among improvements in chronic cough symptoms and taste alterations. All attributes were coded such that the omitted attribute levels corresponded to the reference condition in the study. The model allowed preferences to vary among participants following a normal distribution, thus both a point estimate and SD estimate for each attribute-level parameter distribution were obtained. The resulting log-odds parameter estimates can be interpreted as preference weights that indicate the relative importance of treatment outcomes.17

To evaluate the importance of each attribute over the attribute-level ranges included in the study design, we calculated the difference between the level with the highest relative preference weight and the level with the lowest relative preference weight within each attribute. We then used the sum of all the absolute differences as the denominator to show proportion of the total variation contributed by each attribute, and thus the overall relative importance of each attribute.

Minimum acceptable benefit

Relative preference weights from the RPL model were used to compute minimum-acceptable benefit expressed as the minimum improvement in cough symptoms that respondents would require to offset the negative effect of a change in taste. CIs were estimated using the Krinsky and Robb procedure.18

Latent-class analysis

A logit-based latent-class analysis was performed to identify not only subgroups of participants with similar treatment preferences, but also clinical and survey characteristics that associated with choice patterns. To facilitate comparisons across the classes, preference weight for a bitter, metallic, chalky or oily taste change was set at −1 and all the other attribute levels were rescaled accordingly. Further details about the latent class analysis are provided in online supplemental file 1.


Study cohort

Between 28 January 2022 and 11 April 2022, a total of 502 respondents completed the DCE survey. Thirty participants (6.0%) were excluded from further analyses for always choosing either medication A or medication B in all choice questions, resulting in a final sample size of 472 participants.

The mean (SD) respondent age was 56.9 (14.7) years (table 2). Most participants were white (86.2%), female (79.9%) and without a 4-year college degree (69.1%). In addition to chronic cough, most respondents reported having allergies (70.1%) and gastro-oesophageal reflux disease (65.9%). In addition, 44.1% of participants reported having asthma and 30.5% reported having chronic obstructive pulmonary disease.

Table 2

Characteristics of participants in DCE survey

The majority of participants (347/472, 73.5%) reported having received a specific diagnosis of chronic cough by a physician (table 2). Patients with physician-diagnosed chronic cough were more likely to have increased cough frequency, cough intensity and disruption caused by cough as measured by the Cough Severity Diary (online supplemental table 1). Most respondents reported a current cough (79.4%), with predominant cough timing roughly equally divided among daytime, night-time, and both day and night (table 3). Over 25% of respondents reported coughing 20 times or more per hour during the time of day where their cough was worst. Greater than one-third of respondents reported having seven or more intense cough attacks each week.

Table 3

Characteristics related to cough and prior taste changes in the overall cohort, and stratified by the presence of a self-reported physician diagnosis of chronic cough

One hundred and eighty-nine (40.1%) of respondents reported a prior occurrence of taste alteration from any experience (table 3). Among participants who reported a prior taste alteration, 39.2% experienced decreased ability to taste some kinds of foods, 36.5% experienced an inability to taste some kinds of foods at all and 38.6% experienced a metallic, chalky, oily or bitter taste to foods. The sample was about equally divided between those who thought inability to taste either salty foods or sweet foods would be worse.

Participant preferences for treatment attributes

Participants discriminated well among cough-related attribute levels, indicated by relative preference weights being significantly different from one another (p<0.001, figure 2). In general, participants preferred an improvement in cough symptoms, including less coughing during the night and day along with fewer intense cough attacks. With regard to taste changes, the absence of any taste change was logically most preferred, followed by the inability to taste sweet or salty foods. A bitter, metallic, chalky or oily taste was least preferred. There was no significant difference in relative preference weights between inability to taste salty foods and inability to taste sweet foods (p=0.77). Online supplemental table 2 contains the full set of preference estimates.

Figure 2

Relative preference weights for attributes included in the study. Error bars indicate 95% CIs.

Among the attributes evaluated, the frequency of intense cough attacks was the most important to participants (relative importance weight 33%; 95% CI: 31% to 36%), followed by taste change (relative importance weight 25%; 95% CI 22% to 28%), frequency of night-time coughing (relative importance weight 23%; 95% CI 21% to 26%), and frequency of daytime coughing (relative importance weight 18%; 95% CI 16% to 20%, figure 3). All relative-importance weights were significantly different from each other (p<0.001), with the exception of taste change and frequency of night-time coughing (p=0.44), indicating these attributes were of similar importance.

Figure 3

Relative importance weights for attributes included in the study. Error bars indicate 95% CIs. All relative-importance weights were significantly different from each other (p<0.001), with the exception of taste change and frequency of night-time coughing (p=0.44).

Participant trade-offs among treatment attributes

Participant willingness to make trade-offs was assessed by estimating a minimum improvement in cough symptoms needed in order for respondents to tolerate a change in taste. To accept a bitter, metallic, chalky or oily taste change, participants required either: (1) reduction in night-time cough frequency from 20 to 0 times per hour along with a reduction in daytime cough from 20 to 18 (95% CI 14 to 23) times per hour; (2) reduction in daytime cough frequency from 20 to 0 times per hour and a reduction in night-time cough from 20 to 13 (95% CI 9 to 17) times per hour or (3) a reduction in intense cough attacks from seven times per week to twice (95% CI 1 to 3) a week. A less pronounced improvement in cough symptoms was required to accept an inability to taste sweet or salty foods. Specifically, participants required either: (1) a reduction in night-time cough from 20 to 13 (95% CI 11 to 16) times per hour; (2) a reduction in daytime cough from 20 to 12 (95% CI 9 to 16) times per hour or (3) a reduction in intense cough attacks from 7 times per week to 5.6 (95% CI 5 to 6) times a week.

In a sensitivity analysis that was limited to participants who received a diagnosis of chronic cough by a physician, there was no difference in preferences or trade-offs in comparison to the entire cohort (online supplemental figure 1).

Internal validity testing

On average, respondents correctly answered three of four comprehension questions, and generally improved in the later comprehension questions. Fewer than 1 in 20 participants (4.7%, n=22) always selected the treatment alternative for which one specific attribute had the most desirable level (eg, they always chose the treatment with improvement in night-time cough). Specifically, 6 participants (1.2%) always dominated on frequency of night-time coughing, none on frequency of daytime coughing, 12 (2.5%) on frequency of intense cough attacks and 4 (0.8%) on taste change. A sensitivity analysis excluding participants who dominated on an attribute did not change results.

Latent-class analyses

Based on Bayesian information criterion and Akaike information criterion statistics, three latent classes were identified. A three-class latents model also was parsimonious, plausible and decreased classification errors. The model predicted that 62% of participants were best represented by class-1 preferences, 20% by class-2 preferences and 19% by class-3 preferences.

Class-1 preferences are characterised by being most concerned about intense cough attacks and least concerned about potential taste changes (figure 4A). For participants with class-1 preferences, all cough-related attributes were more important than taste changes, and a less pronounced improvement in cough attributes was required to accept taste disturbances. For example, they were willing to accept a treatment that caused a bitter, metallic, chalky or oily taste as long as it reduced either: (1) night-time cough frequency from 20 times to 9 times per hour; (2) daytime cough frequency from 20 to 6 times per hour or (3) intense cough attacks from 7 to only 5 times per week. In contrast, class-2 preferences are characterised by the desire to avoid certain taste changes. For these participants, the possibility of a bitter, metallic, chalky or oily taste was about 4–5 times more important than any cough attribute (figure 4B). Even a treatment that offered a complete resolution of all cough symptoms would not be acceptable if it were accompanied by a bitter, metallic or chalky taste. Class-3 relative preference weights were disordered with low precision and most were not significantly different from zero (figure 4C).

Figure 4

Relative preference weights for participants with class-1 preferences (A), class-2 preferences (B) and class-3 preferences (C). Participants had a 61.7% chance of having class-1 preferences, a 19.6% chance of having class-2 preferences and an 18.7% chance of having class-3 preferences. Error bars indicate 95% CI.

Of the prespecified clinical and survey characteristics examined, only cough frequency, comprehension quiz score, time spent on completing the survey, and smoking status were found to be associated with class membership. Specifically, both class-1 and class-2 membership significantly associated with a higher score on the comprehension quiz and a baseline cough frequency greater than 20 times per hour during the worst time of day (online supplemental table 3). Class-1 membership was also associated with spending more time on completing the survey. In contrast, class-3 membership significantly associated with performing poorly on the comprehension quiz questions and spending less time on completing the survey. Participants with class-3 preferences also were more likely to be current smokers and less likely to have a baseline cough frequency greater than 20 times per hour. Most clinical characteristics, including measures of cough severity, did not vary significantly between participants’ with relatively high probability of having class-1, class-2 or class-3 preferences (online supplemental table 4).


Patients with refractory chronic cough currently have limited therapeutic options. Recent trials demonstrate P2X3 receptor antagonist gefapixant offers a significant reduction in cough frequency, but taste-related adverse events are common. We evaluated participants’ willingness to accept trade-offs between improvement in chronic cough symptoms and risk of taste-alteration treatment-related side effects. Our findings revealed that participants with chronic cough were willing to accept some taste disturbances in exchange for improved efficacy of chronic cough treatments. Specifically, participants would be willing to tolerate a bitter, metallic, chalky or oily taste change, which was the least preferred taste disturbance, if a cough treatment resulted in either (1) elimination of night-time cough along with a slight reduction in daytime cough; (2) elimination of daytime cough along with a pronounced reduction in night-time or (3) reduction in intense cough attacks from 7 to 2 times per week. Notably, a less pronounced improvement in cough symptoms was required to accept an inability to taste sweet or salty foods.

Prior work has demonstrated that chronic cough has a negative impact on quality of life,19 however, few studies have evaluated relative importance of different cough-related attributes. A strength of this study is the incorporation of qualitative interviews to inform selection of attributes and attribute levels that influence treatment preferences in patients with chronic cough for the DCE survey. Through the DCE, we found that the frequency of intense cough attacks was the most important cough-related attribute to participants, followed by night-time coughing and then daytime coughing. Previous studies reporting on patient experience with chronic cough have described intense cough attacks as prolonged explosive coughing either after a single breath or between several breaths.20 21 Patients have characterised the loss of control associated with cough attacks as well as adverse sequelae after their occurrence, including dizziness, fatigue and chest tightness.22 23 As the first DCE in chronic cough, the current findings add to prior qualitative research by evaluating the extent to which participants are willing to trade improvements in cough attacks for side effects, underscoring their relative importance.

The importance of nocturnal cough compared with daytime cough was somewhat surprising, as objectively measured cough frequency in patients with chronic cough is lower at night than during the day.24 In fact prior work suggests most episodes of nocturnal cough occur while patients are already awake,25 and that sleep inhibits cough through mechanisms such as decreases in cough reflex sensitivity.26 However, studies evaluating the impact of chronic cough on patients report increased sleep disruption and fatigue.19 27 Additionally, nocturnal cough frequency correlates with cough-related quality of life.24 This DCE supports assessment of patient perceived impact on sleep in future cough-targeted clinical trials given the relative importance to participants.

This study identified two distinct preference patterns across participants with chronic cough. Participants with class-1 preferences, prioritised all cough-related attributes, whereas participants with class-2 preferences, prioritised the avoidance of a bitter, metallic, chalky or oily taste change. Interestingly, the proportion of participants with class-2 preferences in this study was similar to the proportion of clinical trial participants who discontinued gefapixant due to taste related adverse events.7 Somewhat surprisingly, a participant’s personal cough frequency or severity did not influence the likelihood of having class-1 versus class-2 preferences. Additionally, there was also no significant difference in responses to cough impact questions derived from the Cough Severity Diary among participants with class-1 versus class-2 preferences. Notably, current smokers were more likely to belong to class-3, whose treatment preferences were less-pronounced, and thus less likely to contribute to overall study results. The heterogeneity in preferences emphasises the importance of considering each individual patient’s perspective in clinical decision-making. Each patient may have a unique set of preferences that cannot be inferred based on demographic or clinical characteristics, and benefit–risk trade-offs will need to be made on an individual basis in the clinic.

This study has several limitations. First, the study design did not limit enrolment to participants with refractory or unexplained chronic cough, and we relied on a participant’s self-reported chronic cough experience. However, about 80% of participants were experiencing a current cough and about 75% had been given a specific diagnosis of chronic cough by a physician. Furthermore, a sensitivity analysis limiting the cohort to participants who had received a diagnosis of chronic cough by a physician produced consistent results. In addition, more pronounced treatment preferences were noted in participants with a baseline cough frequency greater than 20 times per hour. It is thus likely that the preferences identified in this study primarily reflect those of patients with refractory chronic cough. Second, while patients are likely very familiar with their cough, even patients who have experienced taste disturbance in the past may not be familiar with the types of taste alterations evaluated in this survey. In addition, blockade of P2X3 receptor activation could cause many types of taste disturbances,28 29 not all of which were included in this study. However, our DCE focused on the most severe types of taste disturbance experienced by clinical trial participants. Third, the attribute list is not comprehensive, and other attributes relevant to taste disturbance (ie, duration or reversibility), cough (ie, control or duration of cough attacks) or medication cost that all may be important to participants were not included. While including every attribute was not possible, the completion of qualitative interviews to guide DCE development helped to ensure attributes most important to patients were included. Finally, treatment choices in hypothetical settings may not fully correspond to actual behaviour, which is mediated by many factors, including clinical practice and health insurance coverage. However, recent research has shown that in some circumstances, DCEs are able to predict actual choices.30

There have been calls from regulatory agencies to place greater emphasis on the patients’ perspective, including quantitative measures of patients’ treatment preferences.31 DCEs provide important information about attributes that are most valued by patients and the willingness of patients to accept trade-offs between efficacy and side effects. This study provides essential insights into preferences of patients with chronic cough and demonstrates that participants with chronic cough are willing to accept some taste alteration for improved efficacy of chronic cough treatments.

Data availability statement

Deidentified patient data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by Duke IRB Pro00106225. Participants gave informed consent to participate in the study before taking part.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors AS had full access to all of the data in the study and takes full responsibility for the overall content as the guarantor. AS, HD, KG, TC, JS and RJ contributed to the design of the study. J-CY and RJ were involved in data acquisition and in analysis of the data. AS is the guarantor. All authors contributed to interpretation of the data and development of the manuscript. All authors approved the final version.

  • Funding This research was conducted by Duke University with funding provided by Merck Sharp & Dohme, a subsidiary of Merck & Co., Rahway, New Jersey, USA (no award/grant number). All listed authors met authorship criteria established by the International Committee of Medical Journal Editors.

  • Competing interests JS and HD were employees of Merck & Co. at the time this work was completed. No other authors have conflicts of interest.

  • Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.