Chronic cough affects an estimated five to ten percent of adults in the US, and for years there has been no FDA-approved treatment for the condition. Despite billions invested in drug development, promising candidates have stumbled at the regulatory finish line — and some researchers argue that how trials measure cough may be as much a factor as the drugs themselves.
The most prominent recent example is gefapixant, a P2X3 receptor antagonist developed by Merck for refractory or unexplained chronic cough. The FDA's Pulmonary-Allergy Drug Advisory Committee voted 12 to one against approving the drug, largely due to the clinical benefit shown versus placebo, which analysts described as marginal. Both pivotal trials found only a small decrease in cough frequency with gefapixant versus placebo from baseline, and regulators questioned whether such small effects conferred clinically meaningful improvements. The drug was ultimately rejected twice by the FDA, though it has since received approval in Japan and Europe.
The debate over gefapixant has sharpened a broader conversation in cough research: whether the endpoints routinely used in trials are sensitive enough to detect real therapeutic effects — or whether they introduce enough noise to obscure them.
The measurement problem
Joe Brew, a data scientist who has studied cough endpoints in clinical trials, identified three core limitations with current approaches. The first is reliance on patient-reported outcomes. While patient perception remains clinically relevant, research suggests these measures correlate poorly with objective cough frequency — making them imprecise instruments for detecting drug effects.
The second is the short window over which cough is typically measured. "Cough is highly variable hour to hour and day to day across all severity levels," Brew told DDN. "Research shows that at least four days, and ideally seven, are needed to establish an individual's true cough frequency. But, historically, cough has only been measured in 24-hour snapshots." That variability makes it difficult to establish a reliable baseline or to distinguish a true treatment signal from normal day-to-day fluctuation.
The third issue is more subtle: the possibility that monitoring devices themselves alter behavior. Brew argues that bulky or intrusive recording equipment — wires, adhesives, and the like — may introduce the Hawthorne effect, where patients change their behavior simply because they know they are being observed, confounding the very measurements the trial is designed to capture.
Designing better endpoints
What would a more fit-for-purpose cough endpoint look like? Brew's view is that continuous, objective monitoring over multiple days or weeks — rather than episodic 24-hour snapshots — is essential for capturing real-world cough patterns with enough precision to detect meaningful treatment effects. Devices need to be comfortable and unobtrusive enough to sustain high adherence over that period, and the data they produce should be analyzable not just for average cough frequency but for patterns over time, including dose-response relationships.
Sponsors validating these tools, Brew said, should draw on established frameworks including the FDA's guidance on digital health technologies for remote data acquisition and the DiMe V3+ framework, which address analytical validation, clinical relevance, and usability.
There are regulatory barriers to navigate. There is currently no FDA-cleared automated cough counter — existing cleared devices record audio of cough sounds rather than count them automatically. That distinction has required sponsors and regulators to work together to define appropriate validation standards, though Brew noted there is growing optimism in the field that the FDA is prepared to clear the first regulated automated cough counter.
What's at stake for the next wave
Timing matters. A new generation of cough therapeutics, including GSK's camlipixant, is now in clinical development, and the endpoint decisions made at the design stage will shape whether those trials can detect efficacy that may actually be there.
High placebo responses remain a persistent challenge in chronic cough trials, and there are not yet reliable ways to predict which patients will respond to placebo. Some trial designs are exploring whether objective, continuous monitoring — paired with efforts to help patients understand normal symptom variability — can reduce that effect, though Brew acknowledged that work is still ongoing.
For sponsors, the implication is practical: Endpoint selection is not just a methodological detail but a trial design decision with downstream regulatory consequences. "Continuous endpoints allow sponsors to plan for smaller and less expensive studies," Brew said, noting that the pattern has been demonstrated in other indications. "Innovative sponsors are keen to add digital endpoints as exploratory endpoints early in their development process. And when a signal in meaningful cough dynamics is observed, the endpoint is upgraded to secondary or primary."
